[go: up one dir, main page]

WO2018009202A1 - Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules - Google Patents

Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules Download PDF

Info

Publication number
WO2018009202A1
WO2018009202A1 PCT/US2016/041373 US2016041373W WO2018009202A1 WO 2018009202 A1 WO2018009202 A1 WO 2018009202A1 US 2016041373 W US2016041373 W US 2016041373W WO 2018009202 A1 WO2018009202 A1 WO 2018009202A1
Authority
WO
WIPO (PCT)
Prior art keywords
score
cell
cells
discrete
criteria
Prior art date
Legal status (The legal status 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 status listed.)
Ceased
Application number
PCT/US2016/041373
Other languages
English (en)
Inventor
Holger Lange
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Flagship Biosciences LLC
Original Assignee
Flagship Biosciences LLC
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
Application filed by Flagship Biosciences LLC filed Critical Flagship Biosciences LLC
Priority to PCT/US2016/041373 priority Critical patent/WO2018009202A1/fr
Publication of WO2018009202A1 publication Critical patent/WO2018009202A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system

Definitions

  • This invention generally relates to medical imaging; and more particularly, to the analysis of microscopic images from tissue sections.
  • One type of a quantitative scoring scheme is based on a classification of the cells into four ranked categories: 0, 1+, 2+ and 3+.
  • the cells are counted per cell classification category and a discrete ranked score is determined by applying thresholds to the percentages of cells for those four cell classification categories.
  • the IHC HER2 scoring scheme described in Wolff et. al. ⁇ See References is an example of such a quantitative scoring scheme.
  • Cells are classified into the categories: 0, 1+, 2+ and 3+ based on the combination of two cell features, membrane staining intensity and membrane completeness, according to Table 1.
  • Thresholds T(c) of 10%, 10% and 30% are defined corresponding to the inverse cumulative percentages ICP(c).
  • the scores S(c) are defined as 0, 1+, 2+ and 3+ and are associated with the satisfaction of the threshold criteria ICP(c) > T(c), as expressed in Eq. 3. Note that in the case of the IHC HER2 scoring scheme, the cell classification categories and the scores use the same ranked categories.
  • FIGs. 1(A-C) illustrate an example of the IHC HER2 scoring scheme.
  • FIG.1A illustrates the cells (circles) in an image of a tissue section, which are color-coded according to their classification (0 - blue, 1+ - yellow, 2+ - orange and 3+ - red).
  • FIG. IB shows the percentages of cells for the ranked cell classification categories.
  • FIG.1C shows the inverse cumulative percentages for the ranked cell classification categories.
  • the highest cell classification category, where the inverse cumulative percentage is equal to or higher than the threshold is 2+, illustrated by the green arrow.
  • the score, corresponding to the cell classification category 2+ is 2+.
  • Discrete scores like the one provided by the IHC HER2 scoring scheme, provide a classification into clinically-relevant categories, but make it hard to identify borderline cases and to provide a more precise and accurate assessment. While discrete scoring schemes seem to be appropriate for a subjective human evaluation and interpretation, sophisticated image analysis programs that objectively detect the cells on entire tissue sections and quantify the expression of biomarkers can leverage the use of continuous scoring schemes to provide more precise and accurate assessments.
  • the IHC HER2 scoring scheme exhibits an already rather complex cell classification schemes, as it evaluates two cell features, which are still apparently related to the expression of a single biomarker.
  • Using sophisticated image analysis programs that allow characterizing multiple cell features at the same time enables the use of more complex cell classification schemes based on multiple cell features to provide more precise and accurate assessments.
  • Continuous scoring schemes can be developed by expansion of already well- known and discrete scoring schemes that are based on cell classifications. New scoring schemes can be devised that rely on complex cell classification schemes incorporating multiple cell features.
  • the lack of precision and accuracy of existing discrete scoring schemes for the assessment of biomarker expressions in tissue sections can be overcome by using continuous scoring schemes based on cell classifications, in particular when sophisticated image analysis programs are used to detect and characterize the cells and to quantify the biomarker expressions.
  • the invention can be used to extend existing discrete scoring schemes to continuous scoring schemes.
  • the invention can also be used to create new continuous scoring schemes. Using the cell classification as the basis for the scoring scheme provides a simple abstraction from the cell features. It enables the use of complex cell classification schemes based on multiple cell features in the scoring scheme.
  • FIG.1A illustrates cells in an image of a tissue section, which are color-coded according to their classification (0 - blue, 1+ - yellow, 2+ - orange and 3+ - red).
  • FIG. IB shows the percentages of cells for the ranked cell classification categories.
  • FIG.1C shows the inverse cumulative percentages for the ranked cell classification categories
  • FIG.2A shows the inverse cumulative percentages for the ranked cell classification categories in accordance with a continuous scoring scheme.
  • FIG.2B illustrates how the normalized measurements of the criteria are mapped to provide a continuous score.
  • a key aspect to creating a continuous scoring scheme is to use or create a discrete scoring scheme and to provide a formula that provides the continuous values between the discrete scores.
  • One type of a discrete scoring scheme is based on a classification of the cells into a number of ranked categories. The cells are counted per cell classification category and a discrete ranked score is determined by applying thresholds to the percentages of cells in those cell classification categories.
  • the cell classification can be based on a single cell feature (e.g. nuclei staining intensity) or multiple cell features (e.g. membrane staining intensity and membrane completeness) depending on the application.
  • Sophisticated image analysis programs allow characterizing multiple cell features at the same time and even to multiplex cell features across different tissue sections. Any of those cell features, including a characterization of the cell morphology, cell neighborhood, which includes a characterization of the tissue morphology, and the expression of multiple biomarkers, which can include different types of expressions (e.g. protein, gene and mRNA) using different acquisition systems (e.g. brightfield or fluorescence) or just different expressions (e.g. HER2, ER and PR), can be included in a computer-assisted scoring scheme.
  • a single cell feature e.g. nuclei staining intensity
  • multiple cell features e.g. membrane staining intensity and membrane completeness
  • Sophisticated image analysis programs allow characterizing multiple cell features at the same time and even to multiple
  • Any cell classification scheme that allows mapping one or multiple cell features to discrete ranked categories can be used with this method.
  • Computer-assisted scoring schemes can use machine-learning techniques to create cell classifiers that can be based on any number of cell features.
  • the cell classification categories can be defined, depending on the application, from coarse, like the classic cell classification into 0, 1+, 2+ and 3+, to fine, like defining a sampling of the measurements into separate categories (e.g. 0, 1,2,3, ... 255 for 8-bit precision). Without loss of generality, the cell classification categories can be defined as 0, 1,2,3, ... C-1, with C being the number of categories.
  • Thresholds T(c) need to be defined (TBD - to be determined), depending on the application, corresponding to the inverse cumulative percentages ICP(c).
  • Discrete ranked scores S(c) associated with the satisfaction of the threshold criteria ICP(c) > T(c) need to be defined, depending on the application, as expressed in Eq. 6.
  • a formula to provide a continuous scoring for this type of discrete scoring schemes can be based on a function of different criteria, including the confidence into the actual score and the distance to the next higher score.
  • a function that can be used to combine the criteria is the maximum function
  • mapping of the criteria can be done relative to the discrete scores using the half point between the actual score and the next lower score as the anchor to define the mapping intervals corresponding to the different discrete scores. Consequently, the maximum range for the mapping interval would be from the half point between the actual score and the next lower score to the half point between the actual score and the next higher score. Note that when using consecutive numbers for the discrete ranked scores, the rounding of the continuous score will provide the discrete score.
  • the confidence into the actual score can be measured by the difference between the actual percentage of cells ICP(k) and the required percentage of cells to pass the actual threshold T(k) for the actual score S(k). Intuitively, this provides a measure of confidence by how much the actual threshold was passed. If the actual percentage of cells just barely passed the threshold, the continuous score should be close to the border to the next lower score (e.g. 1.5 if the discrete score is 2 and the next lower discrete score is 1). If 100% of the cells passed the threshold, the continuous score should be a full score (e.g. 2.0).
  • Eq. 7. k is the index that corresponds to the actual discrete score. How k is calculated can be seen in Eq. 10.
  • the distance to the next higher score can be measured by the difference between the actual percentage of cells ICP(k+l) to the required percentage of cells to pass the threshold T(k+1) for the next higher score S(k+1). Intuitively, this provides a measure of how close it is to pass the threshold to the next higher score. If 0% of the cells passed the threshold for the next higher score, the actual score should be close to the border to the next lower score (e.g. 1.5 if the discrete score is 2). If the actual percentage of cells is very close to the theshold for the next higher score, the actual score should be close to the border to the next higher score (e.g. 2.49 if the next higher discrete score is 3).
  • Eq. 8 The calculation of the distance to next higher score using a linear mapping and a linear normalization to the maximum range based on the outlined intuition is shown in Eq. 8.
  • the first discrete score S(0) typically chosen to be 0, is the absolute lowest score and the continuous scoring scheme should also start with this first discrete score S(0).
  • the anchor and the mapping intervals for the different criteria may need to be modified for the first discrete score S(0). Using a confidence in the actual score criteria, this can be thought of as always having 100% confidence into the first score S(0). Consequently, such criteria could be implemented as a constant using the maximum of its mapping interval. Using a distance to the next higher score criteria as outlined in Eq. 8, the mapping could be changed to the interval from S(0) to the half point between S(0) and S(l) as shown in Eq. 9. This has the advantage that any percentage of cells that goes towards passing the threshold for the next higher score S(l) are reflected in the continuous score.
  • mapping intervals for the different criteria may need to be modified for the last discrete score S(C-l) as well. Using a distance to the next higher score criteria, this criteria should not apply to the last discrete score as there is no next higher score.
  • FIGs.2(A-B) illustrate the extension of the discrete IHC HER2 scoring scheme to a continuous scoring scheme.
  • FIG.2A shows the inverse cumulative percentages for the ranked cell classification categories as in FIG. 1.
  • the measurements and the corresponding maximum ranges for the normalization for the two criteria, the confidence in the actual score and the distance to the next higher score, are illustrated for the cell classification category 2+ that determined the discrete score of 2+.
  • the difference between the inverse cumulative percentage of cells ICP(2+) and the threshold T(2+) associated with the actual score, shown as a green arrow, is measured for the confidence into the actual score.
  • the corresponding maximum range shown as a gray bar, goes from the threshold T(2+) to 100%.
  • FIG.2B illustrates how the normalized measurements of the criteria are mapped to provide a continuous score.
  • the anchor for the criteria corresponding to the actual discrete score of 2+ is 1.5.
  • the confidence in the actual score criteria is mapped from 1.5 to 2.0.
  • the distance to the next higher score is mapped from 1.5 to 2.0.
  • the normalized measurements from the two criteria taken from FIG.2A are mapped linearly to the mapping intervals.
  • the maximum of the two criteria is calculated to determine the continuous score of 2.3.
  • FIG.3 shows a few different cell distributions with the corresponding continuous HER2 scores.
  • P(0), P(l+), P(2+) and P(3+) are the percentages of cells per cell classification category. Note that rounding of the continuous HER2 score will yield the discrete HER2 score.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

Un procédé d'établissement d'un schéma de notation continue utilisé pour l'évaluation d'expressions de biomarqueurs dans l'analyse de sections de tissus et d'images numériques de celles-ci est basé sur des classifications de cellules.
PCT/US2016/041373 2016-07-07 2016-07-07 Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules Ceased WO2018009202A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/US2016/041373 WO2018009202A1 (fr) 2016-07-07 2016-07-07 Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2016/041373 WO2018009202A1 (fr) 2016-07-07 2016-07-07 Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules

Publications (1)

Publication Number Publication Date
WO2018009202A1 true WO2018009202A1 (fr) 2018-01-11

Family

ID=60913040

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2016/041373 Ceased WO2018009202A1 (fr) 2016-07-07 2016-07-07 Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules

Country Status (1)

Country Link
WO (1) WO2018009202A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110863935A (zh) * 2019-11-19 2020-03-06 上海海事大学 基于VGG16-SegUnet和dropout的海流机叶片附着物识别方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100136549A1 (en) * 2008-09-16 2010-06-03 Historx, Inc. Reproducible quantification of biomarker expression
US20120076390A1 (en) * 2010-09-28 2012-03-29 Flagship Bio Methods for feature analysis on consecutive tissue sections
US20150004630A1 (en) * 2013-02-25 2015-01-01 Flagship Biosciences, LLC Cell-based tissue analysis
US20150003701A1 (en) * 2011-09-16 2015-01-01 Technische Universitat Berlin Method and System for the Automatic Analysis of an Image of a Biological Sample

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100136549A1 (en) * 2008-09-16 2010-06-03 Historx, Inc. Reproducible quantification of biomarker expression
US20120076390A1 (en) * 2010-09-28 2012-03-29 Flagship Bio Methods for feature analysis on consecutive tissue sections
US20150003701A1 (en) * 2011-09-16 2015-01-01 Technische Universitat Berlin Method and System for the Automatic Analysis of an Image of a Biological Sample
US20150004630A1 (en) * 2013-02-25 2015-01-01 Flagship Biosciences, LLC Cell-based tissue analysis

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110863935A (zh) * 2019-11-19 2020-03-06 上海海事大学 基于VGG16-SegUnet和dropout的海流机叶片附着物识别方法

Similar Documents

Publication Publication Date Title
US7979212B2 (en) Method and system for morphology based mitosis identification and classification of digital images
JP6194791B2 (ja) 画像処理装置及びプログラム
Pike et al. Quantifying receptor trafficking and colocalization with confocal microscopy
EP3796217A1 (fr) Classification automatique d'images plein champ à l'aide de l'apprentissage profond
US9488639B2 (en) Cell-based tissue analysis
US6819787B2 (en) Robust stain detection and quantification for histological specimens based on a physical model for stain absorption
JP4504203B2 (ja) 画像分析に基づいたエストロゲン及びプロゲステロン表現のスコアリング
Raimondo et al. Automated evaluation of Her-2/neu status in breast tissue from fluorescent in situ hybridization images
US9760760B2 (en) Histology recognition to automatically score and quantify cancer grades and individual user digital whole histological imaging device
JP7749461B2 (ja) 距離に基づく組織状態の決定
Mashiah et al. Lymphoma and leukemia cells possess fractal dimensions that correlate with their biological features
Cassel et al. Evaluating a 2D image-based computerized approach for measuring riverine pebble roundness
KR20070068401A (ko) 암의 예후에 대한 마커 후보의 분석 및 최적화를 위한 방법및 컴퓨터 프로그램 제품
CN108140239A (zh) 用于组织辨识的方法和装置
WO2010063570A1 (fr) Procédé et système pour identifier des roches
WO2005076197A2 (fr) Procede et systeme d'identification de la mitose sur la base de la morphologie et de classification d'images numeriques
US9412006B1 (en) Continuous tissue analysis scoring scheme based on cell classifications
CN101299242B (zh) 一种人体肤色检测中阈值的确定方法及装置
CN109978849A (zh) 基于数字病理图像的基线细胞确定方法、装置及可读介质
JP7359163B2 (ja) 判別装置、細胞塊の判別方法、及びコンピュータプログラム
WO2018009202A1 (fr) Schéma de notation continue d'analyse de tissu sur la base de classifications de cellules
JP2005296352A (ja) 白髪の評価方法
JP6156137B2 (ja) 画像処理装置及びプログラム
CN116797597B (zh) 一种基于三阶段全网络的脊柱退行性疾病全检测方法及系统
US10482314B2 (en) Automatic calculation for ploidy classification

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16908320

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16908320

Country of ref document: EP

Kind code of ref document: A1