+

WO2005076197A2 - Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques - Google Patents

Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques Download PDF

Info

Publication number
WO2005076197A2
WO2005076197A2 PCT/US2005/003311 US2005003311W WO2005076197A2 WO 2005076197 A2 WO2005076197 A2 WO 2005076197A2 US 2005003311 W US2005003311 W US 2005003311W WO 2005076197 A2 WO2005076197 A2 WO 2005076197A2
Authority
WO
WIPO (PCT)
Prior art keywords
cells
cell
mitotic
pixels
digital image
Prior art date
Application number
PCT/US2005/003311
Other languages
English (en)
Other versions
WO2005076197A3 (fr
Inventor
Abhijeet S. Gholap
Gauri A. Gholap
C.K.V. Rao
Sanford H. Barsky
Madhura Vipra
Suhas Manmantrao Patil
Prithviraj Jadhav
Jayant Abhyankar
Original Assignee
Bioimagene, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US10/938,314 external-priority patent/US20050136509A1/en
Priority claimed from US10/966,071 external-priority patent/US20050136549A1/en
Application filed by Bioimagene, Inc. filed Critical Bioimagene, Inc.
Publication of WO2005076197A2 publication Critical patent/WO2005076197A2/fr
Publication of WO2005076197A3 publication Critical patent/WO2005076197A3/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/457—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by analysing connectivity, e.g. edge linking, connected component analysis or slices
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/60—Extraction of image or video features relating to illumination properties, e.g. using a reflectance or lighting model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/698—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30242—Counting objects in image

Definitions

  • This invention relates to digital image processing. More specifically, it relates to a method and system for morphology based mitosis identification and classification of digital images. BACKGROUND OF THE INVENTION
  • Pathologists use a number of properties in deciding the nature of a cell.
  • a fundamental aspect of histopathology has been the recognition that the morphological appearance of a tumor can be correlated with a degree of malignancy.
  • areas of histopathology such as a diagnosis of breast carcinoma, does not give enough information for the referring medical clinician to make decisions about patient prognosis and treatment. Therefore manual and automated scoring and grading systems used by pathologists have been developed which provide additional information to medical clinicians.
  • One of these automated scoring and grading systems includes considering mitotic activity of cells.
  • Mitosis is a process that facilitates the equal partitioning of replicated chromosomes into two identical groups. Mitosis is a last stage of cell cycle during which cells divide into two cells, hi a typical animal cell, mitosis can be divided into four principal stages: (1) "Prophase:” where cell chromatin, diffuse in interphase, condenses into chromosomes. Each chromosome has duplicated and now consists of two sister chromatids.
  • the nuclear envelope breaks down into vesicles; (2) "Metaphase:” where the chromosomes align at the equitorial plate and are held in place by microtubules attached to the mitotic spindle and to part of the centromere; (3) "Anaphase:” where the centromeres divide. Sister chromatids separate and move toward the corresponding poles; and (4) Telophase: where the daughter chromosomes arrive at the poles and the microtubules disappear. The condensed chromatin expands and the nuclear envelope reappears. The cytoplasm divides, the cell membrane pinches inward ultimately producing two daughter cells (e.g., "Cytokinesis").
  • mitotic count is one of the most important variables in a grading system used for the prognosis of certain cancers including breast cancer. Histological grading has been one of the most important parameters in the determination of the prognosis of breast cancer.
  • One of the important score of cancer grading systems is an evaluation of a Mitotic index of a tissue sample. As is known in the art a "Mitotic index" is an indication of a proliferative activity of a tumor.
  • a Mitotic Activity Index (“MAI”) is a useful and reproducible prognostic indicator for many cancers including invasive breast cancer.
  • the MAI has been defined as the total number of mitoses counted in ten consecutive high-power fields (e.g., objective, x40; numeric aperture, .75; field diameter, 450 microns), in an area subjectively determined to have most cellular activity at the periphery of the tumor.
  • the prognostic value of four methods used to assess mitotic activity in invasive breast cancer was compared in 4-microns-thick hematoxylin-eosin (H&E)- stained sections of 186 primary invasive breast cancer patients. These were the MAI, the random MAI (rMAI), the Mitosis per Volume (M/V) Index, and the random M/V Index (rM/V Index).
  • the rMAI was defined as the total number of mitotic figures counted in 10 random fields through the whole outlined tumor at x400 magnification. All four methods checked had additional prognostic value to tumor size and lymph node status MAI, however, produced the best results, confirming importance of MAI in prognosis of a carcinoma.
  • Mitotic counts differ depending upon the area of tumors analyzed. Margin with the proliferative area give the best results. Mitosis occurs through four successive stages, which are pro-phase, metaphase, anaphase and telophase.
  • a mitosis score is assessed in the peripheral areas of the neoplasm and not in the sclerotic central zone.
  • the neoplasm is scanned at intermediate magnification to determine the area in which mitoses are most abundant (usually areas of poor tubule formation where cells are arranged in sheets or large nests). Only definite mitotic figures are counted with care to avoid non-mitotic nuclei including pyknotic nuclei in the count.
  • the pathologist should be well-trained in the grading of tumors by the S-B-R system.
  • two pathologists can be used and the results obtained by each compared for consistency (Robbins, P., et al., Human Pathology. 26(8):873 (1995)).
  • Tumors can be graded histopathologically on many different bases. As mentioned above, for malignant breast tumors, grading systems such as S-B-R are preferred because they provide objective values of malignancy grade.
  • the pathologist using the S-B-R system looks to three structural characteristics when grading tumors: (1) nuclear pleomorphism; (2) mitotic index; and (3) the ability of the tumor to form tubular, glandular or capillary formations, i.e., ducto glandular differentiation (see, Le Doussal, supra).
  • Tumors are graded by each criterion separately with 1 being the most normal (differentiated) and 3 the most aberrant (undifferentiated). The scores of the three criteria are added for a final tumor grade.
  • the scores can range from 3-5 (well differentiated) to 6-7 (moderately differentiated) and 8-9 (poorly differentiated).
  • Another method is the Nottingham modified criteria of Bloom and Richardson. See Bloom, H. J. G. and Richardson, W. W., Br. J. Cancer 9: 359-377 (1957).
  • This tumor-grading method was based on histological features of tubule formation, nuclear pleomorphism, and mitotic activity, and points were assigned for each category accordingly.
  • the overall tumor grade was the sum total of scores between 3-9.
  • Tumors with poorly differentiated phenotypes (8-9 points) are likely to have less or no tubular structures, irregular and large nuclei, and high mitotic counts.
  • Tumors with moderately (6-7 points) or well differentiated (3-5 points) phenotypes may have definite tubule formation, moderate outlines of epithelial cell shapes and uniformity of nuclear chromatin, and low mitotic indexes.
  • Mitotic activity is graded as follows as per the Nottingham grading system, how many mitotic figures (all four phases) does the pathologist see in 10 high power (400X magnification) fields.
  • Score 1 ⁇ 4 mitoses per square mm. Score 2- 4-7 mitoses per square mm. Score 3: >7 itoses per square mm.
  • Contesso's method of scoring of mitoses is quicker and easier to perform especially on small biopsies (e.g., core biopsies). At least twenty high power fields of the same area as stated above are assessed and scored as is illustrated in Table 3.
  • Score 1 No field contains more than 1 mitosis.
  • Score 2 Two mitoses present in any one HPF.
  • Score 3 Three or more mitoses present in any one HPF.
  • neoplasm whichever is the system followed, accuracy of the detection of mitotic count is most essential.
  • An overall grade of neoplasm is determined by adding individual score of the three separate parameters, tubules, nuclei and mitoses. The grading of the neoplasm has a very important role to play in the treatment and prognosis of the patient.
  • Mitosis occurs through four successive stages, which are pro- phase, metaphase, anaphase and telophase.
  • a mitosis score is typically assessed in peripheral areas of the neoplasm and not in a sclerotic central zone.
  • the neoplasm is typically scanned at intermediate magnification with an optical microscope to determine an area in which mitoses are most abundant (e.g., usually areas of poor tubule formation where cells are arranged in sheets or large nests). Only definite mitotic figures are counted with care to avoid non-mitotic nuclei including pyknotic nuclei in the count.
  • a digital image typically includes an array, usually a rectangular matrix, of pixels.
  • Each "pixel” is one picture element and is a digital quantity that is a value that represents some property of the image at a location in the array corresponding to a particular location in the image.
  • the pixel values typically represent a "gray scale" value.
  • Pixel values for a digital image typically conform to a specified range.
  • each array element may be one byte (i.e., eight bits). With one-byte pixels, pixel values range from zero to 255. In a gray scale image a 255 may represent absolute white and zero total black (or visa- versa).
  • Color images consist of three color planes, generally corresponding to red, green, and blue (RGB). For a particular pixel, there is one value for each of these color planes, (i.e., a value representing the red component, a value representing the green component, and a value representing the blue component). By varying the intensity of these three components, all colors in the color spectrum typically may be created.
  • One type of commonly examined two-dimensional digital images is digital images made from RGB values. Such digital images are commonly used to analyze biological samples including a determination of certain knowledge of medical conditions for humans and animals. For example, digital images are used to determine cell proliferate disorders such as cancers, etc. in humans and animals.
  • Mitosis index for diagnosing cancer.
  • the neoplasm is typically scanned at intermediate magnification with an optical microscope to determine an area in which mitoses are most abundant (e.g., usually areas of poor tubule formation where cells are arranged in sheets or large nests). Only definite mitotic figures are counted with care to avoid non-mitotic nuclei including pyknotic nuclei in the count.
  • H/E stained tissue used for determining digital saliency can also be used for identification and classification of mitosis.
  • H/E stained tissue with a fluorescent signal based mitosis count typically requires stacking image planes in three dimensions to get a focused image. Otherwise, the small fluorescent signals far below the surface of the tissue will give weak, blurred ring of fluorescent signals. Manual method used is time consuming and prone to error including missing areas of the slide including mitotic cells.
  • LSDCAS Large Scale Digital Cell Analysis System
  • LMSTAR S.A. a French-based, company designs, manufactures and markets automated digital imaging systems for Life Sciences Research and Clinical departments in Cytogenetics, Pathology, Cytology, Functional genomic, Drug Development and Validation.
  • IMSTAR launched PATHFINDERTM automated, and cost effective Image Cytometer, associated with a number of Application Software Analysis modules to facilitate and speed up research and diagnostics. This system provides detection of cell proliferation within breast cancer cells.
  • CompuCyte Corporation of Cambridge, Massachusetts has a product for Cell Cycle and DNA Content Analysis.
  • the total amount of DNA per cell is stoichiometrically determined to obtain cell cycle distributions.
  • one of the morphometric features obtained for segmented nuclei, analysis is directly correlated with condensation of chromatin in nuclei and can be used to differentiate interphase cells from mitotic cells.
  • U.S. Patent No. 6,605,432 entitled “High-throughput methods for detecting DNA methylation,” that issued to Tim Hui-Ming Huang teaches “a method of method of hybridization, differential methylation hybridization (DMH) for high throughput methylation analysis of multiple CpG island loci.
  • DMH utilizes nucleic acid probes prepared from a cell sample to screen numerous CpG dinucleotide rich fragments affixed on a screening array. Positive hybridization signals indicate the presence of methylated sites. Methods of preparing the hybridization probes and screening array are also provided.”
  • Also disclosed are methods for diagnosing whether a subject has mild dysplasia, moderate dysplasia, Type A severe dysplasia, Type B severe dysplasia, cervical squamous cell carcinoma, or poorly-differentiated cervical squamous cell carcinoma by determining an expression pattern for PML in an epithelial tissue sample from the subject; dete ⁇ nining an expression pattern for nuclear bodies in the epithelial tissue; determining SUMO-1 colocalization; and determining whether the expression pattern for PML, the expression pattern for nuclear bodies, and the SUMO-1 colocalization of the epithelial tissue sample is consistent with expression patterns expected for mild dysplasia, moderate dysplasia, Type A severe dysplasia, Type B severe dysplasia, cervical squamous cell carcinoma, or poorly-differentiated cervical squamous cell carcinoma.”
  • the invention provides oncology tissue microarrays.
  • the microarrays comprise a plurality of cell and/or tissue samples, each sample representing a different type of cancer.
  • each sample represents a different stage of cancer
  • samples are ordered on the substrate of the microarray into groups according to common characteristics of the patients from whom the samples are obtained. By dividing tissue samples on the substrate into different groupings representing different tissue types, subtypes, histological lesions, and clinical subgroups, the microarrays according to the invention enable ultra-high- throughput molecular profiling.”
  • Luminance parameters such as intensity, etc. from a digital image of a biological sample (e.g., tissue cells) to which a chemical compound has been applied are analyzed. Mitotic cells present in the field of view are automatically identified and classified. The method and system may improve the prognosis and selection of appropriate therapy and prediction of therapeutic outcome based on mitotic cell identification and counting.
  • FIG. 1 is a block diagram illustrating an exemplary automated digital image based mitosis detection and classification system
  • FIG. 2 is a block diagram illustrating the four phases of mitosis illustrated with digital photographs
  • FIG. 3 is a flow diagram illustrating an automated method for automated biological sample analysis
  • FIG. 4 is a block diagram illustrating an automated method for mitosis analysis
  • FIG. 5 is a flow diagram illustrating method for pre-processing digital images to identify mitotic cells
  • FIG. 6 is a flow diagram illustrating an exemplary method for determining a minimum intensity of plural pixels in a digital image to identify mitotic cells
  • FIG. 7 is a flow diagram illustrating a method for contrast enhancement of a digital image to identify mitotic cells
  • FIG. 8A is a block diagram illustrating an original digital image comprising plural cells.
  • FIG. 8B is a block diagram with a histogram of blue, red and green color plane intensity values calculated from the original digital image of FIG. 8 A;
  • FIG. 9A is a block diagram illustrating the original digital image of
  • FIG. 8A which has been contrast enhanced
  • FIG. 9B is a block diagram 80 with a histogram of blue, red and green intensity values calculated from the contrast enhanced digital image of FIG. 9 A;
  • FIG. 10 is a flow diagram illustrating a method for cell object segmentation;
  • FIG. 11 is a block diagram illustrating the effect of tresholding to locate mitotic cells
  • FIG. 12 is a flow diagram illustrating a method for cell object segmentation
  • FIG. 13 is a flow diagram illustrating a method for cell object segmentation
  • FIG. 14 is a flow diagram illustrating a method determining a convex hull part of a mitotic cell boundary in a digital image
  • FIG. 15 is a block diagram of neighborhood masks used to determine a convex hull part in a mitotic cell boundary
  • FIG. 16 is a block diagram illustrating an example of convex hull part in a mitotic cell boundary
  • FIG. 17A is a block diagram illustrating mitotic cells of FIG. 8 A manually identified by a pathologist
  • FIG. 17B is a block diagram illustrating mitotic cells of FIG. 8 A automatically identified Method 44 of FIG. 4;
  • FIG. 18 is a block diagram illustrating an exemplary flow of data in the automated digital image based mitosis detection and classification system
  • FIG. 19 is a flow diagram illustrating a method for determining a gradient within a cell nucleus.
  • FIG. 20 a flow diagram illustrating a method for determining boundary irregularity for a cell nucleus.
  • FIG. 1 is a block diagram illustrating an exemplary automated digital image based mitosis detection and classification system 10.
  • the exemplary system 10 includes one or more computers 12 with a computer display 14 (one of which is illustrated).
  • the computer display 14 presents a windowed graphical user interface ("GUI") 16 with multiple windows to a user.
  • GUI windowed graphical user interface
  • the system 10 may optionally include a microscope or other magnifying device (not illustrated in FIG. 1).
  • the system 10 further includes a digital camera 18 (or analog camera) used to provide plural digital images 20 in various digital images or digital data formats.
  • One or more databases 22 include biological sample information in various digital images or digital data formats.
  • the one or more database 22 may also include raw and processed digital images and may further include knowledge databases created from automated analysis of the digital images 20, report databases and other types of databases as is explained below.
  • the one or more databases 22 may be integral to a memory system on the computer 12 or in secondary storage such as a hard disk, floppy disk, optical disk, or other non-volatile mass storage devices.
  • the computer 12 and the databases 22 may also be connected to an accessible via one or more communications networks 24.
  • the one or more computers 12 may be replaced with client terminals in communications with one or more servers, or with personal digital/data assistants (PDA), laptop computers, mobile computers, Internet appliances, one or two-way pagers, mobile phones, or other similar desktop, mobile or hand-held electronic devices.
  • PDA personal digital/data assistants
  • the communications network 24 includes, but is not limited to, the Internet, an intranet, a wired Local Area Network (LAN), a wireless LAN (WiLAN), a Wide Area Network (WAN), a Metropolitan Area Network (MAN), Public Switched Telephone Network (PSTN) and other types of communications networks 24.
  • LAN Local Area Network
  • WiLAN wireless LAN
  • WAN Wide Area Network
  • MAN Metropolitan Area Network
  • PSTN Public Switched Telephone Network
  • the communications network 24 may include one or more gateways, routers, or bridges.
  • a gateway connects computer networks using different network protocols and/or operating at different transmission capacities.
  • a router receives transmitted messages and forwards them to their correct destinations over the most efficient available route.
  • a bridge is a device that connects networks using the same communications protocols so that information can be passed from one network device to another.
  • the communications network 24 may include one or more servers and one or more web-sites accessible by users to send and receive information useable by the one or more computers 12.
  • the one ore more servers may also include one or more associated databases for storing electronic information.
  • the communications network 24 includes, but is not limited to, data networks using the Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Internet Protocol (IP) and other data protocols.
  • TCP Transmission Control Protocol
  • UDP User Datagram Protocol
  • IP Internet Protocol
  • TCP provides a connection-oriented, end-to-end reliable protocol designed to fit into a layered hierarchy of protocols which support multi-network applications.
  • TCP provides for reliable inter-process communication between pairs of processes in network devices attached to distinct but interconnected networks.
  • IPF Internet Engineering Task Force
  • RRC Request For Comments
  • UDP provides a connectionless mode of communications with datagrams in an interconnected set of computer networks.
  • UDP provides a transaction oriented datagram protocol, where delivery and duplicate packet protection are not guaranteed.
  • IETF RFC- 768 the contents of which incorporated herein by reference.
  • IP is an addressing protocol designed to route traffic within a network or between networks. IP is described in IETF Request For Comments (RFC)-791, the contents of which are incorporated herein by reference. However, more fewer or other protocols can also be used on the communications network 19 and the present invention is not limited to TCP/UDP/IP.
  • the one or more database 22 include plural digital images 20 of biological samples taken with a camera such as a digital camera and stored in a variety of digital image formats including, bit-mapped, joint pictures expert group (JPEG), graphics interchange format (GIF), etc.
  • JPEG joint pictures expert group
  • GIF graphics interchange format
  • the present invention is not limited to these digital image formats and other digital image or digital data formats can also be used to practice the invention.
  • the digital images 20 are typically obtained by magnifying the biological samples with a microscope or other magnifying device and capturing a digital image of the magnified biological sample (e.g., groupings of plural magnified cells, etc.).
  • An operating environment for the devices of the exemplary system 10 include a processing system with one or more high speed Central Processing Unit(s) (“CPU”), processors and one or more memories.
  • CPU Central Processing Unit
  • processors and one or more memories.
  • CPU Central Processing Unit
  • acts and symbolically represented operations or instructions include the manipulation of electrical signals by the CPU or processor.
  • An electrical system represents data bits which cause a resulting transformation or reduction of the electrical signals or biological signals, and the maintenance of data bits at memory locations in a memory system to thereby reconfigure or otherwise alter the CPU's or processor's operation, as well as other processing of signals.
  • the memory locations where data bits are maintained are physical locations that have particular electrical, magnetic, optical, or organic properties corresponding to the data bits.
  • the data bits may also be maintained on a computer readable medium including magnetic disks, optical disks, organic memory, and any other volatile (e.g., Random Access Memory (“RAM”)) or non-volatile (e.g., Read-Only Memory (“ROM”), flash memory, etc.) mass storage system readable by the CPU.
  • RAM Random Access Memory
  • ROM Read-Only Memory
  • the computer readable medium includes cooperating or interconnected computer readable medium, which exist exclusively on the processing system or can be distributed among multiple interconnected processing systems that may be local or remote to the processing system.
  • sample includes cellular material derived from a biological organism. Such samples include but are not limited to hair, skin samples, tissue samples, cultured cells, cultured cell media, and biological fluids.
  • tissue refers to a mass of connected cells (e.g., central nervous system (CNS) tissue, neural tissue, or eye tissue) derived from a human or other animal and includes the connecting material and the liquid material in association with the cells.
  • biological fluid refers to liquid material derived from a human or other animal.
  • sample also includes media containing isolated cells.
  • One skilled in the art may determine the quantity of sample required to obtain a reaction by standard laboratory techniques. The optimal quantity of sample may be determined by serial dilution.
  • biological component include, but not limited to nucleus, cytoplasm, membrane, epithelium, nucleolus and stromal.
  • medical diagnosis includes analysis and interpretation of the state of tissue material in a biological fluid. The interpretation includes classification of tissue sample as “benign tumor cell” or “malignant tumor cell”. Interpretation also includes quantification of malignancy.
  • Mitosis is a process of cell division, which results in the production of two daughter cells from a single parent cell.
  • the daughter cells are identical to one another and to the original parent cell.
  • FIG. 2 is a block diagram 26 illustrating the four phases of mitosis illustrated with digital photographs.
  • mitosis In a typical animal cell, mitosis can be divided into four principal stages illustrated in Table 4.
  • Prophase 28 The chromatin, diffuse in interphase, condenses into chromosomes. Each chromosome has duplicated and now consists of two sister chromatids. At the end of prophase, the nuclear envelope breaks down into vesicles. 2. Metaphase 30: The chromosomes align at the equitorial plate and are held in place by microtubules attached to the mitotic spindle and to part of the centromere. 3. Anaphase 32: The centromeres divide. Sister chromatids separate and move toward the corresponding poles. 4. Telophase 34: Daughter chromosomes arrive at the poles and the microtubules disappear. The condensed chromatin expands and the nuclear envelope reappears. The cytoplasm divides, the cell membrane pinches inward ultimately producing two daughter cells (phase: Cytokinesis).
  • a digital image 20 typically includes an array, usually a rectangular matrix, of pixels.
  • Each "pixel" is one picture element and is a digital quantity that is a value that represents some property of the image at a location in the array corresponding to a particular location in the image.
  • the pixel values typically represent a gray scale value.
  • Pixel values for a digital image 20 typically conform to a specified range.
  • each array element may be one byte (i.e., eight bits). With one- byte pixels, pixel values range from zero to 255. In a gray scale image a 255 may represent absolute white and zero total black (or visa-versa).
  • Color images consist of three color planes, generally corresponding to red, green, and blue (RGB). For a particular pixel, there is one value for each of these color planes, (i.e., a value representing the red component, a value representing the green component, and a value representing the blue component). By varying the intensity of these three components, all colors in the color spectrum typically may be created.
  • One type of commonly examined two-dimensional digital images 20 are digital images made from biological samples including cells, tissue samples, etc. Such digital images are commonly used to analyze biological samples including a determination of certain know medical conditions for humans and animals. For example, digital images are used to determine cell proliferate disorders such as cancers, etc. in humans and animals.
  • Digital images 20 captured through optical microscopes represent the images seen by a human eye through the microscope.
  • a pathologist can easily identify and distinguish between various phases of mitotic cells and non-mitotic cells, even though there are variations in staining, variations in illumination across a slide or the presence of a mask or an artifact. This is because of experience and knowledge of the domain of the pathologist.
  • a Mitotic nucleus is darker than the other nuclei & background. 2. A Mitotic nucleus is comparatively bigger in size. 3. A Mitotic nucleus has sharp edge angles at the boundary Table 5.
  • FIG. 3 is a flow diagram illustrating a Method 36 for automated biological sample analysis.
  • pre-determined parameters from a digital image of a biological sample to which a chemical compound has been applied are modified to make a set of plural biological objects in the digital image more distinct.
  • plural biological objects of interest are located in the set of plural biological objects.
  • the located biological objects of interest are identified and classified to determine a medical diagnosis or medical conclusion.
  • Method 36 is illustrated with an exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • color parameters from a digital image 20 of a tissue sample including plural cells to which a staining dye has been applied are modified to create a set of plural cell objects in the digital image more distinct.
  • the modifications include, but are limited to, contrast modification and correcting color components.
  • a cell object as used herein includes a cell and its various cell components (e.g., membrane, nuclei, chromosomes, etc.).
  • Step 38 is completed as a preprocessing method.
  • the invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • Pre-processing methods reduce the effect of variations in staining intensity, effect of colored mask and other anamolies. Pre-processing methods typically have two distinct steps (1) Contrast modification of an input digital image based on image statistics; (2) Correcting color components of each pixel.
  • contrast modification in a digital image is referred to the difference in color values between any two given pixels. Color values at a given pixel are independently calculated from Red, Green and Blue components of the given color image. Contrast modification includes detennining an active range of intensities in each of the colors. A histogram of all color planes (e.g., red, green and blue) of the digital image 20 are calculated. The calculated histograms are used to compute a minimum intensity such that, starting from lowest intensity, cumulative pixels up to minimum intensity is equal to about two percent of total pixels in the digital image. An active range intensity range of pixels in the digital image 20 is mapped to a range (zero, 255). All pixels with value less than minimum intensity are also set to zero.
  • a histogram of all color planes e.g., red, green and blue
  • Color mask parameters are calculated from the calculated color planes using peak frequencies in respective calculated histograms. Ratios of the peak values in one or more color planes are used to correct pixels in the digital image 20.
  • plural mitotic cells are located as objects of interest in a set of plural cells in a tissue sample in a digital image.
  • tissue sample image there could be several hundred cells.
  • These cells could be normal epithelial cells, stained epithelial cells, stromal cells or mitotic cells in any one of the mitotic phases 28, 30, 32, 34. Mitotic cells are located as objects of interest and non-mitotic cells are eliminated from further processing.
  • the located mitotic cells are classified to determine a medical diagnosis or medical conclusion.
  • a medical diagnosis may include a diagnosis of N-stage breast cancer
  • the Nottingham grading system or counting a number of mitoses in ten high power fields (HPFs) is used to classify the mitotic cells.
  • HPFs high power fields
  • the present invention is not limited to such an embodiment and other classification system can be used to practice the invention.
  • FIG. 4 is a block diagram illustrating an automated Method 44 for mitosis analysis.
  • luminance values of plural pixels from a digital image of a biological sample to which a chemical compound has been applied are analyzed and adjusted if necessary, to identify plural cells.
  • the identified plural cells are segmented using morphological parameters to remove non-mitotic cells.
  • plural mitotic cells are identified from the segmented plural cells.
  • the plural mitotic cells are classified using a pre-determined classification scheme to create a medical diagnosis or prognosis.
  • Method 44 may be specifically used by pathologists and other medical personnel to automatically analyze a tissue sample for mitotic cells and make a medical diagnosis or prognosis.
  • the present invention is not limited to such an application and Method 44 may also be used for other purposes.
  • Method 44 may also be used for automatically determining diagnostic saliency of digital images for mitotic cells.
  • This method can be used for automatically determining diagnostic saliency of digital images for mitotic cells and includes using one or more filters for evaluating digital images 20. Each filter is designed to identify a specific type of morphological parameter of a mitotic cell.
  • Methods 44 may also be used for automatically quantitatively analyzing biological samples. This method is use for automatically quantitatively analyzing relevant properties of the digital images, and creating interpretive data, images and reports resulting from such analysis.
  • Method 44 further include Step 53 (not illustrated in FIG. 4) to create one or more reports related to the medical diagnosis or prognosis created and present digital image 20 and the one or more types of reports generated for the medical diagnosis or prognosis on the GUI 14.
  • Method 44 achieves the same and in most instances better results than those completed by a human pathologist, but in an automated manner (See, e.g., Table 9).
  • the present invention is not limited this method and other methods can be used to practice the invention.
  • Method 44 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • luminance parameters such as intensity, etc. from a digital image 20 of a biological sample (e.g., tissue cells) to which a chemical compound (e.g., a marker dye) has been applied are analyzed and corrected if necessary to identify plural cells in the digital image.
  • a biological sample e.g., tissue cells
  • a chemical compound e.g., a marker dye
  • Step 46 includes a pre-processing method illustrated with Method 54.
  • the present invention is not limited to such an embodiment and other methods can be used at Step 46 to practice the invention.
  • FIG. 5 is a flow diagram illustrating Method 54 for pre-processing digital images to identify mitotic cells.
  • a minimum pixel intensity is determined in plural color planes independently in one or more areas of interest in a digital image.
  • color components of selected pixels in the one or more determined areas of interest are corrected using the determined minimum intensity.
  • Method 54 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • the pre-processing Method 54 includes, but is not limited to, at least two distinct steps: (1) Determining a minimum pixel intensity of a digital image based on digital image statistics at Step 56; and (2) correcting color components of selected pixels in the one or more determined areas of interest using the determined minimum intensity at Step 58.
  • the present invention is not limited to these steps and more, fewer or other steps can also be used to practice the invention.
  • a digital image is considered "high contrast” if its luminosity levels range from a minimum value (e.g., zero) to a maximum value (e.g., 255). In the case of low contrast images, this range could be as small as zero to 50, for example, or range from 100 to 150.
  • FIG. 6 is a flow diagram illustrating an exemplary Method 60 for determining a minimum intensity of plural pixels in a digital image to identify mitotic cells.
  • Step 62 histograms of pixel values are calculated for each color plane independently in a digital image.
  • Step 64 a minimum intensity of an active range of pixels is determined using a distribution of pixel values from the calculated histogram in each of the color planes independently.
  • Method 60 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • a threshold value is used for independently determining a mimmum intensity in each color plane using the calculated cumulative frequencies.
  • the threshold value is two percent
  • the present invention is not limited to this threshold value and other threshold values may be use to practice the invention.
  • pixel values in each color plane are mapped such that pixel values in a first range ⁇ minimum intensity to maximum intensity ⁇ are mapped to a second range ⁇ zero to maximum intensity ⁇ to correct the color components and contrast enhance the digital image.
  • Contrast enhancement or difference between pixels level is increased by setting all pixels below minimum intensity level to zero, keeping and keeping maximum intensity in each color plane the same.
  • FIG. 7 is a flow diagram illustrating a Method 66 for contrast enhancement of a digital image to identify mitotic cells.
  • Step 68 pixel values corresponding to peaks in histograms in color planes within segmented areas of interest within a digital image are determined.
  • Step 70 ratios of pixels are calculated in selected color planes within the segmented areas of interest. The color planes used are determined by the chemical compound (e.g., staining such as H/E staining) applied to the biological tissue sample.
  • pixel color values are modified with the segmented areas of interest using pre-defined equations.
  • Method 66 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • color mask parameters are calculated from the red, green and blue color planes using peak frequencies in respective histograms for the color planes.
  • ratios of peak frequency values are calculated in the appropriate color planes.
  • the color planes used are determined by the chemical compound applied to the biological tissue sample.
  • FIG. 8 A is a block diagram illustrating an original digital image 74 comprising plural cells.
  • FIG. 8B is a block diagram with a histogram 76 of blue 76', red 76" and 76'" color plane intensity values calculated from the original digital image 74.
  • H/E staining is used so the red and blue color planes are used to calculate ratios at Step 70.
  • segmented objects in areas of interest such as mitotic cells
  • nuclei are blue in color when stained with H/E staining and other stains. If a biological tissue sample was treated with other than H/E staining, then nuclei or other cell components may appear as a different color other than blue and thus other color planes would be used at Step 70.
  • This ratio calculation is based on chemical compound (e.g., a stain such as H/E stain) that has been applied to the biological sample.
  • the invention is not limited to this embodiment and other ratios in other color planes can also be used to practice the invention.
  • Equations (3) and (4) are used to compute the ratios of color plane histogram peaks.
  • the fist color plane is the red color plane and the second color plane is the blue color plane.
  • the red peak and blue peak are respective color component values that have peak frequency as is illustrated in the histogram 76. If the image is predominantly blue colored, then a blue peak will be much larger than a red peak. If the image is predominantly red in color, then red peak will be larger than the blue peak. Histogram 76 illustrates a blue peak 76' much larger than a red peak 76". Equations (5) and (6) illustrate calculating ratios for am image that is primarily blue in color.
  • R f Red color plane peak/ Blue color plane peak (6)
  • Equations (7), (8), and (9) are used to compute modified Red, Green and Blue component values of a pixel in the digital image.
  • R,G, and B are red, green and blue component values of a pixel respectively and R',G', and B' are red, green and blue component values of a modified pixel respectively and CI, C2 and C3 are pre-determined constants.
  • all of the pre-determined constants have a value of two.
  • the present invention is not limited to this embodiment and other constant values can also be used to practice the invention.
  • the pre-determined constants CI, C2 and C3 do not have to be the same value.
  • Equation (7) a contrast in blue plane pixels is increased. If the pixel has blue component value less than the peak of blue plane, the term (C* B —Blue peak) will be less than B. If the pixel has blue component value greater than the peak of blue plane, the term C* B -Blue peak) will be greater than B. Therefore the difference between two pixel values, one greater than peak and the other less than peak will be increased.
  • a multiplication factor "R f " is used to normalize peak intensity of blue color plane with respect to red color plane peak.
  • a minimum condition used in the equation ensures that the Blue component never exceeds a predetermined constant X (e.g., 255) and maximum condition used ensures that the Blue component value never becomes negative.
  • Equation (8) contrast in the Red plane pixels is increased. If the pixel has Red component value less than the peak of Red plane, the term (C* R - Red peak) will be less than R. If the pixel has Red component value greater than the peak of Red plane, the term (C* R - Red peak) will be greater than R. Therefore the difference between two pixels values, one greater than mean and the other less than mean will be increased.
  • a multiplication factor "B f " is used to normalize peak intensity of red color plane with respect to blue color plane peak
  • a minimum condition used in the equation ensures that the Red component never exceeds a predetermined constant X, (e.g., 255) and a maximum condition used ensures that the Red component value never becomes negative.
  • Equation (9) contrast in the Green plane pixels is increased. If the pixel has Green component value less than the mean of Green plane, the term (C* G - Green peak) will be less than G. If the pixel has Green component value greater than the mean of Green plane, the term (C* G - Green peak) will be greater than Pixel Intensity. Therefore the difference between two pixel values, one greater than peak and the other less than peak will be increased.
  • a minimum condition used in the equation ensures that the Green component never exceeds a pre-determined constant X, (e.g., 255) and a maximum condition used ensures that the Green component value never becomes negative.
  • FIG. 9A is a block diagram illustrating a digital image 78 which has been contrast enhanced or corrected using the methods described herein. Note objects of interest of this contrast enhanced digital image 78 are darker than the original image 74 (FIG. 8A).
  • FIG. 9B is a block diagram 80 with a histogram of blue 80', red 80" and green 80'" color intensity values calculated from the contrast enhanced digital image 78. Note how blue 80' color values have been enhanced compared to red 80" color values.
  • the identified plural cells are segmented using mo ⁇ hological parameters to remove non-motitic cells.
  • the mo ⁇ hological parameters include cell size, cell elongation, cell component parallelism, degree cell boundary roughness or smoothness and cell shape including convex or concave shapes.
  • the invention is not limited to such an embodiment and other mo ⁇ hological parameters can also be used to practice the invention.
  • any given tissue sample image there could be several hundred cells of different types. These cells could be normal epithelial cells, stained epithelial cells, stromal cells or mitotic cells in any one of the mitotic phases. Objects of interest including mitotic cells are segmented and non-mitotic cells are deleted from further processing. One step in segmentation is to locate an object of interest via a thresholding operation. Thresholding is carried out on the modified digital image 78.
  • FIG. 10 is a flow diagram illustrating a Method 82 for cell object segmentation.
  • a threshold operation is applied to eliminate cells in the plural identified cells that correspond to non-mitotic cells.
  • boundaries of mitotic cells are determined in the remaining plural identified cells.
  • Method 82 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • mitotic cells are segmented using thresholding.
  • thresholding Using H/E staining, mitotic cells are blue in color and are of different size and shapes than non-mitotic cells.
  • the red and blue color planes of the modified image 78 are used in deciding whether a given pixel belongs to a mitotic cell are not.
  • a threshold value of 20% of a total range of grayscale value of 50 is used for segmenting mitotic cells.
  • the present invention is not limited to such a threshold value and other threshold values can also be used to practice the invention.
  • mitotic cells are cropped to accurate cell boundaries.
  • the mitotic cell boundaries may not be actual mitotic cell boundaries.
  • An area of interest being identified as a mitotic cell is smaller than the actual cell due to hard thresholds being used.
  • Mitotic cells smaller than about 200 pixels in size are deleted and the remaining objects are labeled. Labeled objects are cropped till a relaxed boundary condition is met. Another different threshold is used for cropping mitotic cells to an accurate boundary.
  • FIG. 11 is a block diagram 88 illustrating the effect of tresholding to locate objects of interest 90 (e.g., mitotic cells).
  • objects of interest 90 e.g., mitotic cells.
  • the objects of interest 90 appear darker than the surrounding biological objects.
  • luminosity of pixels in all color planes is used instead of red plane value to segment mitotic cells.
  • FIG. 12 is a flow diagram illustrating a Method 92 for cell object segmentation.
  • Step 94 plural luminosity values are determined for plural color planes at plural (x,y) positions in a digital image.
  • mitotic cells in the plural identified cells are cropped until a dete ⁇ nined luminosity value is more than a pre-determined percentage of a total range of luminosity values.
  • Method 92 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • R(x,y), G(x,y) and B(x,y) are red, green and blue plane values at (x,y) position in the digital image 20 respectively.
  • the present invention is not limited to such an embodiment and other constants can be used to practice the invention.
  • mitotic cells are cropped until a determined luminosity value I(x,y) is more than a pre-determined percentage (e.g., 20%) of a total range of luminosity values (e.g., a value of 50).
  • both thresholding and luminosity values are used to segment the identified plural cells.
  • FIG. 13 is a flow diagram illustrating a Method 98 for cell object segmentation.
  • a first area is calculated for a cell using thresholding operation on plural pixels in the cell.
  • a second area is calculated for the cell using plural luminosity values calculated for the cell.
  • a test is conducted to determine whether a ratio of the ((first area) / (second area)) is more than a pre-determined segmenting value, and if so, at Step 106, the cell is identified as a non-mitotic cell and removed from the identified plural cells. If at Step 104 the ratio is less than the pre-determined segmenting value, then at Step 108 the cell is identified as a mitotic cell and not removed from the identified plural cells.
  • Method 98 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • a first area, AREAl calculated gives an area value which is based on a thresholding operation using red color plane values as was described above in Equations (3-9).
  • a second area, AREA2 is calculated gives an area value based luminosity values of pixels as was describe above for Equation (10) instead of red color plane values.
  • the ratio, AREAl/ AREA2 is more than the pre-determined segmenting value (e.g., 0.55), then at Step 102, the cell object is identified as non-mitotic. In one embodiment, a pre-determined value of 0.55 is used.
  • the present invention is not limited to such an embodiment and other pre-determined segmenting can also used to practice the invention. If at Step 104 the ratio is less than the pre-determined segmenting value, then at Step 108 the cell is identified as a mitotic cell and not removed from the identified plural cells.
  • plural mitotic cells are identified from the segmented plural cells.
  • plural filters are used to identify mitotic cells.
  • Table 6 illustrates exemplary filters applied to identify mitotic cells of interest.
  • the present invention is not limited to the steps illustrated in Table 6 and other filters can also be used to practice the invention.
  • Size based filters any given tissue sample image, there could be several hundred cells, but there typically will be only a small number of mitotic cells. Mitotic cells are larger than normal cells and are often larger than 15 microns. At standard resolution of 40X, a 15-micron object corresponds to about 300 pixels. Artifacts and blood vessels are very big compared to mitotic cells. Large artifacts and blood vessels are filtered by limiting the size to 100 microns or 10000 pixels at 40X resolution. Size based filters are used to filter away such large artifacts. One more stage of size-based filtering then is used to filter mitotic cell objects.
  • mitotic cell size is always above average cell size. Mitotic cell objects less than an average size are removed for further processing. While utilizing the information that mitotic cells are larger than normal cells, pathologists seldom use numbers attached to these terms. Terms like "large” are always relative.
  • a size of mitotic cells objects in terms of pixels is used for identification as belonging to a Telophase 34 mitotic activity.
  • large artifacts are filtered by limiting object size to 1500 pixels at 40X resolution, i such an embodiment, if any two objects have size more than 1500 pixels, then this pair of cell objects cannot belong to Telophase 34 mitotic cell.
  • the present invention is not limited to such an embodiment and other object sizes can also be used to practice the invention.
  • Elongation ratio based filters At two stages of the evolution, namely, Prophase 28 and Metaphase 30, mitotic cells exhibit elongated dumb bell shape.
  • An "elongation ratio" is defined as a ratio of major axis over minor axis. This ratio is used in determining an object shape factor of a cell.
  • a center of a cell object is determined.
  • a major axis is determined by computing a length of the object in all directions passing through a center of the object.
  • a length and angle of a maximum length line passing through a center of the object gives major axis and angle of major axis.
  • Minor axis is considered to be pe ⁇ endicular to the major axis.
  • the minor axis is determined based on line passing through the center of the object and pe ⁇ endicular to the major axis.
  • other calculations can also be used to calculate an elongation ration and this embodiment.
  • an exemplary elongation ratio of 1.5 is used to eliminate circular objects is used.
  • the present invention is not limited to this ratio and other access ratios can also be used to practice the invention.
  • Elongation ratios are checked. If an object has an aspect ratio much larger than one, this means its length is twice the width. This is more likely a mitotic cell. On the other hand, if the aspect ratio is one or less, then the object is circular or nearly circular in shape. These circular objects are not part of a Telophase 34 mitotic cell, hi one embodiment, if any of these two objects have aspect ratio value less than 1.5, then this pair of objects cannot belong to Telophase 34 mitotic cell.
  • Mitotic cells are typically very long compared to normal cells, in particular mitotic cells in Metaphase 30 a length of major axis will be very high.
  • Non-mitotic cells are filtered using length of a major axis.
  • an exemplary hard threshold value of 85 pixels is used for length of major axis to filter non-mitotic cells.
  • the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • Size of a mitotic cell image does not vary on re-segmentation with a relaxed threshold. Therefore a ratio of size of an object with hard threshold and relaxed threshold can also be used to filter non-mitotic cells.
  • Another important characteristic of mitotic cells in Telophase 34 is that two components of a mitotic cell are well separated and are parallel. So a parallelism property of two identified mitotic cells is used to identify mitotic cells.
  • Parallelism based filter Cell objects that are not parallel because of a difference in proximity size, width, length and elongation are detected. Parallelism based filters are used.
  • a proximity of two separate objects is used to determine if the pair belongs to Telophase 34 mitotic cell or not.
  • Telophase 34 two separate mitotic cells are still in close proximity and are not migrated to other parts of tissue Euclidian distance between two mitotic cells is used to estimate the proximity, hi order to facilitate variations in sizes of mitotic cells, in one embodiment a predetermined proximity threshold limit of four times a maximum length L of the two objects is used. A pair of mitotic cells farther than this pre-determined proximity threshold limit cannot belong to Telophase 34.
  • the present invention is not limited to such an embodiment and other pre-determined proximity threshold limits can also be used to practice the invention.
  • a difference in a length "L" of cell objects is used to identify
  • Telophase 34 mitotic cells. Both parts of mitotic cells in Telophase 34 are supposed to be a same length as these are created from a single cell.
  • a pair of cell objects can be considered as two components in Telophase 34 if the difference in their length is less than a pre-determined length threshold of a cell object having maximum length in a pair under consideration, hi one embodiment, the pre-determined length threshold of 30% is used for difference in length to eliminate non-Telophase 34 mitotic cells.
  • the present invention is not limited to such an embodiment and other predetermined length thresholds can also be used to practice the invention.
  • a difference in a width "W" of cell objects is used to identify
  • Telophase 34 mitotic cells Both parts of mitotic cells in Telophase 34 are supposed to be of the same width as these are created from a single cell.
  • a pair of objects can be considered as two components in Telophase 34 if a difference in their width is less than a pre-determined percentage of an object having maximum width in a pair of objects under consideration, hi one embodiment, a pre-determined width threshold of 30% is used for difference in width to eliminate non-Telophase 34 mitotic cells.
  • a pre-determined width threshold of 30% is used for difference in width to eliminate non-Telophase 34 mitotic cells.
  • the present invention is not limited to such an embodiment and other predetermined width thresholds can also be used to practice the invention.
  • a difference in the size "S" of objects is used to identify Telophase 34 mitotic cells. Both parts of mitotic cells in Telophase 34 are supposed to be of the same size as these are created from a single cell. A pair of objects can be considered as two components in Telophase 34 if a difference in their size is less than a predetermined percentage of the object having maximum size in the pair under consideration. In one embodiment, a pre-determined size threshold of 30% is used to determine a difference in size to eliminate non-Telophase 34 mitotic cells. However, the present invention is not limited to such an embodiment and pre-determined size thresholds can also be used to practice the invention. [00171] A number of parameters are used in determining if two given mitotic cells are parallel and close enough to be classified as Telophase 28. A first parameter Pi is illustrated in Equation (11):
  • Ei is an elongation ratio
  • Li is a length of a major axis
  • Thetai is an angle of major axis with a X-axis
  • Wi is a width of objecti at the center of object
  • Si is a size of objecti.
  • Equation (12) A second parameter P 2 is illustrated in Equation (12):
  • E 2 is an elongation ratio
  • L 2 is a length of a major axis
  • Theta 2 is an angle of major axis with a X-axis
  • W 2 is a width of object2 at the center of objecti
  • S S 2 is a size of objecti.
  • Cell objects one and two can be classified as mitotic cells if the conditions listed in Table 7 are satisfied for standard images at 40X resolution.
  • the present invention is not limited to the conditions in Table 7 and more, fewer or other conditions can also be used to practice the invention.
  • Si ⁇ 1500 and S 2 ⁇ 1500 • Difference between an angle of major axis, Theta., and Theta 2 is less than 25 degrees.
  • Distance between the centers of the two objects is less than four times a maximum of L. ⁇ and l_ 2 , where the distance is a Euclidean distance between the centers of the objects.
  • a difference in of major axis of two objects is used to decide the parallel nature of mitotic cells in Telophase 34.
  • Thetai indicates an angle of a major axis of object one makes with an x-axis.
  • Theta 2 indicates a angle major axis of object two makes with the x-axis.
  • the difference in angles will be negative if Thetais less than Theta . Therefore an absolute value is used.
  • a threshold angle of 25 degrees is used for difference in angle to eliminate non-parallel mitotic cell pairs.
  • the present invention is not limited to this threshold angle and other angles can also be used to practice the invention.
  • Convex hull based filter Convex hulls are also features of a mitotic cell. Normal cells are concave in shape. A ratio of convex hull pixels is measured to a size of the object in order to distinguishing between mitotic cells and dying cells (e.g., crimped). Dying cells, which are crimped, have convex hulls, but this convex hull ratio will be very large.
  • convex hull As is known in the art convex hull have been used for boundary description in digital images. However, most convex hulls known in the art use a change in slope for detecting a convex hull boundary. For example, R.C. Gonzalez, and R.E. Woods in "Digital Image processing," Pearson Education, 2003 pp. 653- 655 describe a method using a change in slope for detecting a convex hull boundary.
  • a neighborhood based operator is used instead of a change in slope to detect a convex hull boundary. Neighborhood based operations on binary images are faster and more efficient compared to sequential operations like finding slope at a pixel. All identified mitotic cells are considered as two level image objects and analyzed for a convex hull part, hi one embodiment, neighborhood operations are implemented using a pre-determined size neighborhood mask (e.g., 3 x 3).
  • FIG. 14 is a flow diagram illustrating a Method 110 for determining a convex hull part of a mitotic cell boundary in a digital image.
  • Step 112 plural neighborhood masks are centered around plural pixels on a boundary of a cell.
  • Step 114 a test is conducted to determine if the pixels belong to a convex hull boundary of the cell. If the cell boundary is a convex hull of the cell at Step 114, a convex hull factor is calculated for the cell at Step 116.
  • the calculated convex hull factor is used to determine if the cell is a mitotic cell.
  • Method 110 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • Step 110 plural pixel neighborhood masks are centered around every pixel on a boundary of a cell object to determine if the pixels belong to a convex hull of the cell object
  • FIG. 15 is a block diagram 120 of exemplary neighborhood masks, 122, 124, 126, 128 used to determine a convex hull part in a mitotic cell boundary.
  • the present invention is not limited to these neighborhood masks and other neighborhoods masks can be used to practice the invention.
  • a neighborhood mask (e.g., 3 x 3) is centered around every pixel on a boundary of a cell object to determine if the pixel belongs to a convex part of the cell object.
  • a cell object in this context is a two level image, where a value one implies it is a pixel on a mitotic cell and a value zero implies it is a pixel not on the mitotic cell.
  • a pixel with value zero having three neighbors all with value one is identified as a pixel in the convex hull of the object.
  • a pixel is identified as pixel on convex hull part of boundary if it satisfies any of the neighborhood masks 122-128.
  • the present invention is not limited to this embodiment and other embodiments can also be used to practice the invention.
  • FIG. 16 is a block diagram 130 illustrating an example of convex hull part 122 in a mitotic cell boundary.
  • the convex hull part 132 is darker in color.
  • Step 116 a ratio of pixels is used satisfying convex hull condition over the cell object size.
  • H/ be the convex hull factor defined as is illustrated in Equation (13).
  • Hf (number of pixels on a convex hull) / (number of pixels in a cell) (13)
  • a cell obj ect has H f , in a range of about 0.05 to 0.70, then the cell object is a mitotic cell. If Hf, is less than about 0.05, it means that the cell object is concave in nature (i.e., a non-mitotic or "normal" cell). If Hf, is more than about 0.70 then the cell object is has a very large hull part (i.e., the cell is crimped and dying).
  • Hf is less than about 0.05, it means that the cell object is concave in nature (i.e., a non-mitotic or "normal" cell). If Hf, is more than about 0.70 then the cell object is has a very large hull part (i.e., the cell is crimped and dying).
  • Hf values i.e., the cell is crimped and dying.
  • Shape Base Filter During Metaphase 30 and Telophase 34 stages of the cell division, mitotic cells exhibit elongated dumb bell shape. This shape factor is reliable means of classifying mitotic cells. A dumb bell shape is used to identify mitotic cells. A ratio of a major axis over minor axis is used in determining the object shape factor of a mitotic cell. In order to compute elongation ratio, a center of the object is determined first by finding the mean values of x and y coordinates of all pixels in the object. A major axis is determined by computing a length of the chords in all directions passing through the center of the object. A length and angle of a maximum length chord passing through the center of the object gives major axis and angle of major axis.
  • a minor axis is considered to be pe ⁇ endicular to the major axis.
  • the minor axis is determined based on line passing through a center of the object and pe ⁇ endicular to the major axis.
  • An elongation ratio defined as the ratio of major axis over minor axis is dete ⁇ nined.
  • a predetermined shape threshold of 1.5 is used for the elongation ratio to eliminate circular objects.
  • the present invention is not limited to such an embodiment and other pre-determined shape thresholds can also be used to practice the invention
  • Boundary Smoothness Filter Mitotic cells have different characteristic during Prophase 28 including checkered boundary with a number of thread like extensions. Boundary smoothness factor is used to filter objects that exhibit rough boundaries. Segmented objects which are binary in nature are considered for analysis. Objects under consideration are eroded by one pixel thickness. Subtracting an eroded object from an original object provides an object boundary. It is observed that a standard operation like smoothening digital image by applying low pass filter like Gaussian operator eliminates high frequency variations in a boundary and gives an estimation of boundary roughness. A Gaussian filter is applied to create a smoothened version of the identified object. A boundary of a smoothened object under consideration is determined by eroding the object by one pixel thickness and then subtracting eroded object from the original.
  • a center of gravity of the identified object boundary is determined by finding the sum of all x coordinates of pixels on a boundary "Xsum,” finding the sum of all y coordinates of pixels on the boundary "Ysum.”
  • An X coordinate of the center of gravity can be calculated by dividing Xsum by a number of pixels on the boundary.
  • a Y coordinate of the center of gravity can be calculated by dividing Ysum by the number of pixels on the boundary.
  • Boundary roughness is calculated based on a difference in distance between the center of gravity and pixels on the two boundaries, namely one without Gaussian smoothening and the other with Gaussian smoothening. In one embodiment, 360 pixels on boundary are considered for calculating boundary roughness.
  • the present invention is not limited to such an embodiment and more or fewer pixels can be used on a boundary and other boundary calculations can also be used to practice the invention.
  • a difference in distance will be of same sign, either positive or negative for circular objects, but will have several changes in sign for objects with convex hulls.
  • the number of times the difference in distance changes sign indicates a boundary roughness value.
  • a given object is identified as mitotic if any of the following three Equations (16), (17), (18) are satisfied.
  • Ri is a first constant (e.g., 4.0)
  • R 2 is a second constant (e.g., 7.0)
  • R 3 is a third constant (e.g., 10).
  • the present invention is not limited to these constants and other constants can also be used to practice the invention.
  • Prophase 28 mitotic cell exhibits rough boundary at a local level
  • a boundary smoothness factor is determined. Segmented objects which are binary in nature are considered for analysis. Objects under consideration are eroded by one pixel thickness. Subtracting eroded object from the original gives object boundary. Pixels on an object boundary are chain coded using 8- connectivity. Chain coding and 8 -connectivity are extensively covered in the literature including the text book "Digital Image Processing" by Gonzalez R C, and Woods R E, Pearson Education, 2003. Isolated noise on the chain code is eliminated. A difference in chain code between adjacent pixels on the object boundary is calculated. This difference varies between zero and seven. For a linear portion of any curve, this difference between adjacent pixels will be zero. Therefore by calculating the ratio of number of pixels with zero difference in chain code to the total number of pixels on the boundary a measure of boundary smoothness is determined.
  • the plural mitotic cells are classified using a pre-determined classification scheme to create a medical diagnosis or prognosis.
  • the pre-determined classification scheme includes a Nottingham grading system or counting a number of mitoses in ten high power fields (HPFs) that is used to classify the mitotic cells.
  • HPFs high power fields
  • the present invention is not limited to such an embodiment and other pre-determined classification schemes or systems can be used to practice the invention.
  • a medical diagnosis may include a diagnosis of N-stage breast cancer or a medical prognosis such as terminal breast cancer with six months to live.
  • FIG. 17 illustrates one exemplary set of actual results of using Method
  • FIG. 8 A is a block diagram 134 illustrating mitotic cells 136 of FIG. 8A manually identified by a pathologist.
  • Method 44 automatically identified seventeen mitotic cells. There were 17 actual cells undergoing mitosis in FIG. 8A.
  • FIG. 17B is a block diagram 138 illustrating mitotic cells 140 of FIG.
  • Table 8 A automatically identified Method 44 of FIG. 4.
  • Table 8 illustrates an exemplary error rate calculated between FIGS. 17A and 17B.
  • automated Method 44 determined all mitotic cells including many missed by a pathologist performing manual methods.
  • FIG. 18 is a block diagram illustrating an exemplary flow of data 142 in the automated digital image based mitosis detection and classification system 10.
  • Pixel values from a digital image 20 of a biological sample to which a chemical compound has been applied are captured 144 as raw digital images 146.
  • the raw digital images 146 are stored in raw image format in one or more image databases 22.
  • Luminance and mo ⁇ hological parameters from individual biological components within the biological sample are analyzed on the digital image and modifications made to the raw digital images are used to create new biological knowledge 148 using the methods described herein.
  • the new biological knowledge is stored in a knowledge database 150. Peer review of the digital image analysis and life science and biotechnology experiment results is completed 152.
  • a reference digital image database 154 facilitates access of reference images from previous records of life science and biotechnology experiments at the time of peer review. Contents of the reference digital image database 154, information on the biological sample and analysis of current biological sample are available at an image retrieval, reporting and informatics module 156 that displays information on GUI 14. Conclusions of a medical diagnosis or prognosis or life science and biotechnology experiment are documented as one or more reports. Report generation 158 allows configurable fields and layout of the report. New medical knowledge is automatically created and saved.
  • ANN Artificial Neural Networks
  • an ANN based on FIG. 18 is used for training and classifying mitotic cells from plural biological samples analyzed over time.
  • a given cell input object is classified as mitotic or non-mitotic based on three parameters values. These parameters are a gradient variation within a cell object, cell object boundary irregularity and cell object elongation ratio as was discussed above. However, more, fewer or other parameter may also be used to practice the invention.
  • FIG. 19 is a flow diagram illustrating a Method 160 for determining a gradient with a cell nucleus.
  • Step 162 a minimum and maximum gradient in an area of interest of a cell nucleus in a digital image and a histogram of the gradient of corresponding pixels used are calculated.
  • Step 164 a first average gradient across the area of interest is calculated.
  • Step 166 a test is conducted to determine if the calculated maximum gradient is less than a pre-determined gradient or a predetermined number of pixels with a gradient greater then the pre-determined gradient is less than a pre-determined number, then at Step 168, a calculated gradient variation is set to zero. Otherwise at Step 170, a pre-determined number of pixels having a largest gradient are selected.
  • Step 172 a second average gradient variation is determined.
  • Step 174 a calculated gradient is set to (second average gradient - first average gradient).
  • Step 162 Minimum and maximum gradients in an of area interest in a cell nucleus and a histogram of the gradient of corresponding pixels used are calculated.
  • Step 164 a first average gradient, AVGl, across the area of interest is calculated.
  • Step 166 a test is conducted to determine a gradient variation within the cell nucleus.
  • Step 168 if a maximum gradient is less than a pre-determined gradientl (e.g., 20), or a number of pixels with a gradient more than the pre-determined gradientl is less than a pre-determined numberl (e.g., 10), then a calculated gradient variation is set to zero at Step 170.
  • a pre-determined gradientl e.g. 20
  • a pre-determined numberl e.g. 10
  • a second average gradient, AVG2, for selected pixels is calculated at Step 172.
  • a pre-determined number e.g., 20 pixels having largest a gradient are selected. This selection is done based on the histogram of the gradient within nucleus calculated at Step 162.
  • the calculated gradient variation is set to (AVG2- AVGl). This value indicates a variation in gradient within the cell nucleus. This value along with other two features is used by artificial neural network for classifying object as mitotic or non-mitotic.
  • the pre-determined gradientl includes a value of 20
  • the pre-determined numberl includes a value of 10
  • the predetermined number of pixels having a largest gradient includes a value of 20.
  • the present invention is not limited to these values and values can also be used to practice the invention.
  • FIG. 20 a flow diagram illustrating a Method 176 for determining boundary irregularity for a cell nucleus.
  • Step 178 plural pixels on a boundary of cell nucleus in a digital image are identified.
  • Step 180 a center of gravity of the identified plural boundary pixels is calculated.
  • Step 182 a boundary irregularity is calculated using the computer center of gravity and the identified plural boundary pixels.
  • Method 176 is illustrated with one exemplary embodiment. However, the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention. [00208] In such an exemplary embodiment at Step 178, cell nuclei segmented with one or more of the segmenting methods described above which are binary in nature are considered for analysis. A cell nucleus under consideration is eroded by one pixel thickness. Subtracting eroded cell nucleus pixels from the original cell nucleus gives an object boundary.
  • a center of gravity of the identified cell nucleus boundary is calculated by finding a sum of all x coordinates of pixels on the boundary, Xsum, finding a sum of all y coordinates of pixels on the boundary, Ysum.
  • An "x" coordinate of the center of gravity can be calculated by dividing Xsum by a number of pixels on the boundary.
  • a "y" coordinate of the center of gravity can be calculated by dividing Ysum by the number of pixels on the boundary.
  • a boundary irregularity is calculated based on the variation in distance between the center of gravity and pixels on the boundary.
  • pixels on boundary at an interval of 5 degrees are considered for calculating boundary irregularity.
  • This distance will be constant for circular objects, but will have several zero crossings for objects with convex hulls. Zero crossing is with respect to average distance between the center of gravity and pixels on the boundary. The number of times the distance value is making a zero crossing indicates the boundary irregularity value.
  • the present invention is not limited to these degree value or distances and values can also be used to practice the invention.
  • This boundary irregularity is also used by the ANN for classifying object as mitotic or non- mitotic.
  • an ANN with back propagation network with one hidden layer is used to practice the invention using the methods and systems described herein.
  • the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention, hi such an embodiment, the ANN is trained for a maximum of 5000 cycles or a threshold on average error, hi such an embodiment, an average error of 0.01 is used for terminating training.
  • the present invention is not limited to such an embodiment and other embodiments can also be used to practice the invention.
  • the present invention is implemented in software.
  • the invention may be also be implemented in firmware, hardware, or a combination thereof, including software. However, there is no special hardware or software required to use the proposed invention.
  • the methods and system described herein are used to provide an automated medical conclusion or a life science and biotechnology experiment conclusion is determined from the analyzed luminance and morphological parameters of mitotic cells.
  • the method and system is also used for automatically obtaining a medical diagnosis (e.g., a carcinoma diagnosis) or prognosis.
  • the method and system may also be used to provide an automated medical conclusion for new drug discovery and/or clinical trials used for testing new drugs.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

L'invention porte sur un procédé et sur un système d'identification de la mitose sur la base de la morphologie et de classification d'images numériques. Selon cette invention, on analyse et on corrige, si besoin, des paramètres de luminance tels que l'intensité, etc., provenant de l'image numérique d'un échantillon biologique (par exemple, des cellules tissulaires) sur lequel a été appliqué un composé chimique (tel qu'un colorant de marqueur). On analyse également des paramètres morphologiques (tels que taille, rapport d'allongement, parallélisme, rugosité de surface, forme de coques convexes, etc.) à partir de composantes individuelles à l'intérieur de l'échantillon biologique. On procède à une conclusion médicale (type et comptage de cellules mitotiques) ou à une conclusion d'expérimentations biologiques et biotechnologiques à partir des paramètres morphologiques et de luminance analysés. Le procédé et le système de l'invention peuvent être utilisés pour mette au point des applications en vue d'obtenir automatiquement un diagnostic médical (par exemple, un diagnostic de carcinome).
PCT/US2005/003311 2004-01-31 2005-01-31 Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques WO2005076197A2 (fr)

Applications Claiming Priority (6)

Application Number Priority Date Filing Date Title
US54076704P 2004-01-31 2004-01-31
US60/540,767 2004-01-31
US10/938,314 2004-09-10
US10/938,314 US20050136509A1 (en) 2003-09-10 2004-09-10 Method and system for quantitatively analyzing biological samples
US10/966,071 US20050136549A1 (en) 2003-10-30 2004-10-15 Method and system for automatically determining diagnostic saliency of digital images
US10/966,071 2004-10-15

Publications (2)

Publication Number Publication Date
WO2005076197A2 true WO2005076197A2 (fr) 2005-08-18
WO2005076197A3 WO2005076197A3 (fr) 2005-10-27

Family

ID=34841732

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2005/003311 WO2005076197A2 (fr) 2004-01-31 2005-01-31 Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques

Country Status (1)

Country Link
WO (1) WO2005076197A2 (fr)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010003041A3 (fr) * 2008-07-03 2010-03-25 Nec Laboratories America, Inc. Détecteur de figure mitotique et système de compteur ainsi que procédé de détection et de comptage de figures mitotiques
CN102156988A (zh) * 2011-05-27 2011-08-17 天津大学 一种细胞分裂序列检测方法
US8417015B2 (en) 2007-08-06 2013-04-09 Historx, Inc. Methods and system for validating sample images for quantitative immunoassays
US8428887B2 (en) 2003-09-08 2013-04-23 Ventana Medical Systems, Inc. Method for automated processing of digital images of tissue micro-arrays (TMA)
US8515683B2 (en) 2003-09-10 2013-08-20 Ventana Medical Systems, Inc. Method and system for automated detection of immunohistochemical (IHC) patterns
US8655037B2 (en) 2007-05-14 2014-02-18 Historx, Inc. Compartment segregation by pixel characterization using image data clustering
EP2737435A1 (fr) * 2011-07-27 2014-06-04 Omnyx, Llc Systèmes et procédés de pathologie numérique
EP2756453A1 (fr) * 2011-09-12 2014-07-23 Google, Inc. Système d'amélioration de contenu
US8873833B2 (en) 2012-02-17 2014-10-28 Sony Corporation System and method for effectively performing a scene representation procedure
EP3657432A1 (fr) * 2018-11-23 2020-05-27 Acer Incorporated Procédé de normalisation d'image et dispositif de traitement d'image
CN111292287A (zh) * 2018-12-06 2020-06-16 宏碁股份有限公司 图像正规化方法及图像处理装置
EP3677168A4 (fr) * 2017-08-31 2020-11-04 Osaka University Appareil de diagnostic de pathologie, procédé de traitement d'images et programme associé
CN115359056A (zh) * 2022-10-19 2022-11-18 浙江华诺康科技有限公司 分裂细胞检测方法、装置和计算机设备
US20220414986A1 (en) * 2021-06-23 2022-12-29 Adobe Inc. Segmenting three-dimensional meshes in graphical applications based on detection of elongated shapes
CN116824250A (zh) * 2023-06-27 2023-09-29 重庆邮电大学 一种基于形态学和香农熵的细胞形态分类方法

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2737116C (fr) 2008-09-16 2019-01-15 Historx, Inc. Quantification reproductible de l'expression de biomarqueurs

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
DREZET P M L ET AL: "Automatic mitotic index estimation for the prognostication of breast cancer from histology images" IEE COLLOQUIUM INTELLIGENT METHODS IN HEALTHCARE AND MEDICAL APPLICATIONS, 20 October 1998 (1998-10-20), pages 14-1, XP006502885 *
KATE TEN T K ET AL: "METHOD FOR COUNTING MITOSES BY IMAGE PROCESSING IN FEULGEN STAINED BREAST CANCER SECTIONS" CYTOMETRY, ALAN LISS, NEW YORK, US, vol. 14, 1993, pages 241-250, XP009026418 ISSN: 0196-4763 *
MADACHY R J ET AL: "Image analysis for automatic classification of MITOTIC cervical cells" PROCEEDINGS OF THE ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY, 4 November 1988 (1988-11-04), pages 372-374, XP010074652 *
SUNDBLAD LARS-GORAN ET AL: "The use of image analysis and automation for measuring mitotic index in apical conifer mersitems" JOURNAL OF EXPERIMENTAL BOTANY, OXFORD UNIVERSITY PRESS, GB, vol. 49, no. 327, October 1998 (1998-10), pages 1749-1756, XP002233559 ISSN: 0022-0957 *

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8428887B2 (en) 2003-09-08 2013-04-23 Ventana Medical Systems, Inc. Method for automated processing of digital images of tissue micro-arrays (TMA)
US8515683B2 (en) 2003-09-10 2013-08-20 Ventana Medical Systems, Inc. Method and system for automated detection of immunohistochemical (IHC) patterns
US8655037B2 (en) 2007-05-14 2014-02-18 Historx, Inc. Compartment segregation by pixel characterization using image data clustering
US8417015B2 (en) 2007-08-06 2013-04-09 Historx, Inc. Methods and system for validating sample images for quantitative immunoassays
WO2010003041A3 (fr) * 2008-07-03 2010-03-25 Nec Laboratories America, Inc. Détecteur de figure mitotique et système de compteur ainsi que procédé de détection et de comptage de figures mitotiques
CN102156988A (zh) * 2011-05-27 2011-08-17 天津大学 一种细胞分裂序列检测方法
EP2737435A4 (fr) * 2011-07-27 2015-04-08 Omnyx LLC Systèmes et procédés de pathologie numérique
EP2737435A1 (fr) * 2011-07-27 2014-06-04 Omnyx, Llc Systèmes et procédés de pathologie numérique
EP2756453A1 (fr) * 2011-09-12 2014-07-23 Google, Inc. Système d'amélioration de contenu
EP2756453A4 (fr) * 2011-09-12 2015-04-08 Google Inc Système d'amélioration de contenu
US8873833B2 (en) 2012-02-17 2014-10-28 Sony Corporation System and method for effectively performing a scene representation procedure
JPWO2019044579A1 (ja) * 2017-08-31 2020-11-05 国立大学法人大阪大学 病理診断装置、画像処理方法及びプログラム
EP3677168A4 (fr) * 2017-08-31 2020-11-04 Osaka University Appareil de diagnostic de pathologie, procédé de traitement d'images et programme associé
US10755393B2 (en) 2018-11-23 2020-08-25 Acer Incorporated Image normalization method and image processing device
EP3657432A1 (fr) * 2018-11-23 2020-05-27 Acer Incorporated Procédé de normalisation d'image et dispositif de traitement d'image
CN111292287A (zh) * 2018-12-06 2020-06-16 宏碁股份有限公司 图像正规化方法及图像处理装置
US20220414986A1 (en) * 2021-06-23 2022-12-29 Adobe Inc. Segmenting three-dimensional meshes in graphical applications based on detection of elongated shapes
US12056823B2 (en) * 2021-06-23 2024-08-06 Adobe Inc. Segmenting three-dimensional meshes in graphical applications based on detection of elongated shapes
CN115359056A (zh) * 2022-10-19 2022-11-18 浙江华诺康科技有限公司 分裂细胞检测方法、装置和计算机设备
CN116824250A (zh) * 2023-06-27 2023-09-29 重庆邮电大学 一种基于形态学和香农熵的细胞形态分类方法

Also Published As

Publication number Publication date
WO2005076197A3 (fr) 2005-10-27

Similar Documents

Publication Publication Date Title
US7979212B2 (en) Method and system for morphology based mitosis identification and classification of digital images
US7760927B2 (en) Method and system for digital image based tissue independent simultaneous nucleus cytoplasm and membrane quantitation
Merino et al. Optimizing morphology through blood cell image analysis
Forero et al. Automatic identification of Mycobacterium tuberculosis by Gaussian mixture models
JP7076698B2 (ja) 画像解析方法、画像解析装置、プログラム、学習済み深層学習アルゴリズムの製造方法および学習済み深層学習アルゴリズム
US20050265588A1 (en) Method and system for digital image based flourescent in situ hybridization (FISH) analysis
Keenan et al. An automated machine vision system for the histological grading of cervical intraepithelial neoplasia (CIN)
US20050136549A1 (en) Method and system for automatically determining diagnostic saliency of digital images
US8068988B2 (en) Method for automated processing of digital images of tissue micro-arrays (TMA)
US20190042826A1 (en) Automatic nuclei segmentation in histopathology images
AU2015265811B2 (en) An image processing method and system for analyzing a multi-channel image obtained from a biological tissue sample being stained by multiple stains
CN111448569B (zh) 存储和检索数字病理学分析结果的方法
EP1470411B1 (fr) Procede de video-microscopie quantitative et systeme associe, et produit de programme logiciel informatique
CN105027165B (zh) 用于数字完整载片的自动化评分的基于组织对象的机器学习系统
WO2005076197A2 (fr) Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques
US9147104B2 (en) Systems and methods for processing low contrast images
EP2050070A2 (fr) Système et procédés destinés à noter des images provenant d'un microréseau tissulaire
WO2004044845A2 (fr) Analyse d'image
US11538261B2 (en) Systems and methods for automated cell segmentation and labeling in immunofluorescence microscopy
WO2007024264A2 (fr) Procede et systeme de quantification simultanee, a base d'images numeriques et independante du tissu du noyau, du cytoplasme et de la membrane
Wang et al. A rule-based computer scheme for centromere identification and polarity assignment of metaphase chromosomes
EP1579366A1 (fr) Evaluation histologique de pleomorphisme nucleaire
WO2005096225A1 (fr) Procede et systeme de detection automatises de profils immunohistochimiques (ihc)
WO2005076216A2 (fr) Procede et systeme pour l'analyse automatique d'hybridation in situ par fluorescence a base d'image numerique
Wang et al. Automated detection and analysis of fluorescent in situ hybridization spots depicted in digital microscopic images of Pap-smear specimens

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
NENP Non-entry into the national phase

Ref country code: DE

WWW Wipo information: withdrawn in national office

Country of ref document: DE

122 Ep: pct application non-entry in european phase
点击 这是indexloc提供的php浏览器服务，不要输入任何密码和下载