Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning
<p><strong>Objective:</strong> Complex phenotypes captured on histological slides represent the biological processes at play in individual cancers but the link to underlying molecular classification has not been clarified or systematised. In colorectal cancer (CRC), histological gr...
| Hlavní autoři: | , , , , , , , , , |
|---|---|
| Médium: | Journal article |
| Jazyk: | English |
| Vydáno: |
BMJ Publishing Group
2020
|
| _version_ | 1826260528784211968 |
|---|---|
| author | Sirinukunwattana, K Domingo-Villanueva, E Richman, S Blake, A Verrill, C Leedham, SJ Wu, C-H Maughan, T Rittscher, J Koelzer, VH |
| author_facet | Sirinukunwattana, K Domingo-Villanueva, E Richman, S Blake, A Verrill, C Leedham, SJ Wu, C-H Maughan, T Rittscher, J Koelzer, VH |
| author_sort | Sirinukunwattana, K |
| collection | OXFORD |
| description | <p><strong>Objective:</strong> Complex phenotypes captured on histological slides represent the biological processes at
play in individual cancers but the link to underlying molecular classification has not been clarified or systematised. In colorectal cancer (CRC), histological grading is a poor predictor of disease progression, and consensus molecular subtypes (CMS) cannot be distinguished without gene expression profiling. We hypothesise that image analysis is a cost-effective tool to associate complex features of tissue organisation with molecular and outcome data and to resolve unclassifiable or
heterogeneous cases. In this study, we present an image-based approach to predict CRC CMS from standard H&E sections using deep learning.</p>
<p><strong>Design:</strong> Training and evaluation of a neural network were performed using a total of n=1,206 tissue
sections with comprehensive multi-omic data from three independent datasets (training on FOCUS trial, n=278 patients; test on rectal cancer biopsies, GRAMPIAN cohort, n=144 patients, and The Cancer Genome Atlas (TCGA), n=430 patients). Ground truth CMS calls were ascertained by matching random forest and single sample predictions from MSclassifier.</p>
<p><strong>Results:</strong> Image-based CMS (imCMS) accurately classified slides in unseen datasets from TCGA
(n=431 slides, AUC=0.84) and rectal cancer biopsies (n=265 slides, AUC=0.85). imCMS spatially resolved intra-tumoural heterogeneity and provided secondary calls correlating with bioinformatic prediction from molecular data. imCMS classified samples previously unclassifiable by RNA expression profiling, reproduced the expected correlations with genomic and epigenetic alterations and showed similar prognostic associations as transcriptomic CMS.</p>
<p><strong>Conclusion:</strong> This study shows that a prediction of RNA expression classifiers can be made from H&E
images, opening the door to simple, cheap and reliable biological stratification within routine workflows.</p> |
| first_indexed | 2024-03-06T19:07:04Z |
| format | Journal article |
| id | oxford-uuid:157de4da-8dc8-4f98-bc7e-16dc946c28e7 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T19:07:04Z |
| publishDate | 2020 |
| publisher | BMJ Publishing Group |
| record_format | dspace |
| spelling | oxford-uuid:157de4da-8dc8-4f98-bc7e-16dc946c28e72022-03-26T10:25:58ZImage-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learningJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:157de4da-8dc8-4f98-bc7e-16dc946c28e7EnglishSymplectic ElementsBMJ Publishing Group2020Sirinukunwattana, KDomingo-Villanueva, ERichman, SBlake, AVerrill, CLeedham, SJWu, C-HMaughan, TRittscher, JKoelzer, VH<p><strong>Objective:</strong> Complex phenotypes captured on histological slides represent the biological processes at play in individual cancers but the link to underlying molecular classification has not been clarified or systematised. In colorectal cancer (CRC), histological grading is a poor predictor of disease progression, and consensus molecular subtypes (CMS) cannot be distinguished without gene expression profiling. We hypothesise that image analysis is a cost-effective tool to associate complex features of tissue organisation with molecular and outcome data and to resolve unclassifiable or heterogeneous cases. In this study, we present an image-based approach to predict CRC CMS from standard H&E sections using deep learning.</p> <p><strong>Design:</strong> Training and evaluation of a neural network were performed using a total of n=1,206 tissue sections with comprehensive multi-omic data from three independent datasets (training on FOCUS trial, n=278 patients; test on rectal cancer biopsies, GRAMPIAN cohort, n=144 patients, and The Cancer Genome Atlas (TCGA), n=430 patients). Ground truth CMS calls were ascertained by matching random forest and single sample predictions from MSclassifier.</p> <p><strong>Results:</strong> Image-based CMS (imCMS) accurately classified slides in unseen datasets from TCGA (n=431 slides, AUC=0.84) and rectal cancer biopsies (n=265 slides, AUC=0.85). imCMS spatially resolved intra-tumoural heterogeneity and provided secondary calls correlating with bioinformatic prediction from molecular data. imCMS classified samples previously unclassifiable by RNA expression profiling, reproduced the expected correlations with genomic and epigenetic alterations and showed similar prognostic associations as transcriptomic CMS.</p> <p><strong>Conclusion:</strong> This study shows that a prediction of RNA expression classifiers can be made from H&E images, opening the door to simple, cheap and reliable biological stratification within routine workflows.</p> |
| spellingShingle | Sirinukunwattana, K Domingo-Villanueva, E Richman, S Blake, A Verrill, C Leedham, SJ Wu, C-H Maughan, T Rittscher, J Koelzer, VH Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title | Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title_full | Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title_fullStr | Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title_full_unstemmed | Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title_short | Image-based consensus molecular subtype classification (imCMS) of colorectal cancer using deep learning |
| title_sort | image based consensus molecular subtype classification imcms of colorectal cancer using deep learning |
| work_keys_str_mv | AT sirinukunwattanak imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT domingovillanuevae imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT richmans imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT blakea imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT verrillc imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT leedhamsj imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT wuch imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT maughant imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT rittscherj imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning AT koelzervh imagebasedconsensusmolecularsubtypeclassificationimcmsofcolorectalcancerusingdeeplearning |