Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning
Abstract Studies have shown that colorectal cancer prognosis can be predicted by deep learning-based analysis of histological tissue sections of the primary tumor. So far, this has been achieved using a binary prediction. Survival curves might contain more detailed information and thus enable a more...
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-09-01
|
| Series: | npj Precision Oncology |
| Online Access: | https://doi.org/10.1038/s41698-023-00451-3 |
| _version_ | 1797454117010931712 |
|---|---|
| author | Julia Höhn Eva Krieghoff-Henning Christoph Wies Lennard Kiehl Martin J. Hetz Tabea-Clara Bucher Jitendra Jonnagaddala Kurt Zatloukal Heimo Müller Markus Plass Emilian Jungwirth Timo Gaiser Matthias Steeg Tim Holland-Letz Hermann Brenner Michael Hoffmeister Titus J. Brinker |
| author_facet | Julia Höhn Eva Krieghoff-Henning Christoph Wies Lennard Kiehl Martin J. Hetz Tabea-Clara Bucher Jitendra Jonnagaddala Kurt Zatloukal Heimo Müller Markus Plass Emilian Jungwirth Timo Gaiser Matthias Steeg Tim Holland-Letz Hermann Brenner Michael Hoffmeister Titus J. Brinker |
| author_sort | Julia Höhn |
| collection | DOAJ |
| description | Abstract Studies have shown that colorectal cancer prognosis can be predicted by deep learning-based analysis of histological tissue sections of the primary tumor. So far, this has been achieved using a binary prediction. Survival curves might contain more detailed information and thus enable a more fine-grained risk prediction. Therefore, we established survival curve-based CRC survival predictors and benchmarked them against standard binary survival predictors, comparing their performance extensively on the clinical high and low risk subsets of one internal and three external cohorts. Survival curve-based risk prediction achieved a very similar risk stratification to binary risk prediction for this task. Exchanging other components of the pipeline, namely input tissue and feature extractor, had largely identical effects on model performance independently of the type of risk prediction. An ensemble of all survival curve-based models exhibited a more robust performance, as did a similar ensemble based on binary risk prediction. Patients could be further stratified within clinical risk groups. However, performance still varied across cohorts, indicating limited generalization of all investigated image analysis pipelines, whereas models using clinical data performed robustly on all cohorts. |
| first_indexed | 2024-03-09T15:32:37Z |
| format | Article |
| id | doaj.art-e094c1a7b01c4b7bb8979efab082563b |
| institution | Directory Open Access Journal |
| issn | 2397-768X |
| language | English |
| last_indexed | 2024-03-09T15:32:37Z |
| publishDate | 2023-09-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Precision Oncology |
| spelling | doaj.art-e094c1a7b01c4b7bb8979efab082563b2023-11-26T12:12:23ZengNature Portfolionpj Precision Oncology2397-768X2023-09-017111210.1038/s41698-023-00451-3Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learningJulia Höhn0Eva Krieghoff-Henning1Christoph Wies2Lennard Kiehl3Martin J. Hetz4Tabea-Clara Bucher5Jitendra Jonnagaddala6Kurt Zatloukal7Heimo Müller8Markus Plass9Emilian Jungwirth10Timo Gaiser11Matthias Steeg12Tim Holland-Letz13Hermann Brenner14Michael Hoffmeister15Titus J. Brinker16Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)School of Population Health, Faculty of Medicine and Health, UNSW SydneyDiagnostic and Research Center for Molecular BioMedicine, Diagnostic & Research Institute of Pathology, Medical University of GrazDiagnostic and Research Center for Molecular BioMedicine, Diagnostic & Research Institute of Pathology, Medical University of GrazDiagnostic and Research Center for Molecular BioMedicine, Diagnostic & Research Institute of Pathology, Medical University of GrazDiagnostic and Research Center for Molecular BioMedicine, Diagnostic & Research Institute of Pathology, Medical University of GrazInstitute of Pathology, University Medical Center Mannheim, University of HeidelbergInstitute of Pathology, University Medical Center Mannheim, University of HeidelbergDepartment of Biostatistics, German Cancer Research Center (DKFZ)Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ)Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ)Digital Biomarkers for Oncology Group, German Cancer Research Center (DKFZ)Abstract Studies have shown that colorectal cancer prognosis can be predicted by deep learning-based analysis of histological tissue sections of the primary tumor. So far, this has been achieved using a binary prediction. Survival curves might contain more detailed information and thus enable a more fine-grained risk prediction. Therefore, we established survival curve-based CRC survival predictors and benchmarked them against standard binary survival predictors, comparing their performance extensively on the clinical high and low risk subsets of one internal and three external cohorts. Survival curve-based risk prediction achieved a very similar risk stratification to binary risk prediction for this task. Exchanging other components of the pipeline, namely input tissue and feature extractor, had largely identical effects on model performance independently of the type of risk prediction. An ensemble of all survival curve-based models exhibited a more robust performance, as did a similar ensemble based on binary risk prediction. Patients could be further stratified within clinical risk groups. However, performance still varied across cohorts, indicating limited generalization of all investigated image analysis pipelines, whereas models using clinical data performed robustly on all cohorts.https://doi.org/10.1038/s41698-023-00451-3 |
| spellingShingle | Julia Höhn Eva Krieghoff-Henning Christoph Wies Lennard Kiehl Martin J. Hetz Tabea-Clara Bucher Jitendra Jonnagaddala Kurt Zatloukal Heimo Müller Markus Plass Emilian Jungwirth Timo Gaiser Matthias Steeg Tim Holland-Letz Hermann Brenner Michael Hoffmeister Titus J. Brinker Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning npj Precision Oncology |
| title | Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| title_full | Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| title_fullStr | Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| title_full_unstemmed | Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| title_short | Colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| title_sort | colorectal cancer risk stratification on histological slides based on survival curves predicted by deep learning |
| url | https://doi.org/10.1038/s41698-023-00451-3 |
| work_keys_str_mv | AT juliahohn colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT evakrieghoffhenning colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT christophwies colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT lennardkiehl colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT martinjhetz colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT tabeaclarabucher colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT jitendrajonnagaddala colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT kurtzatloukal colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT heimomuller colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT markusplass colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT emilianjungwirth colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT timogaiser colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT matthiassteeg colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT timhollandletz colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT hermannbrenner colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT michaelhoffmeister colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning AT titusjbrinker colorectalcancerriskstratificationonhistologicalslidesbasedonsurvivalcurvespredictedbydeeplearning |