Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor
Abstract Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, the AFIP/Miettinen classification depends heavily on mitotic counts, which is laborious and sometimes inconsisten...
| Main Authors: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-07-01
|
| Series: | npj Precision Oncology |
| Online Access: | https://doi.org/10.1038/s41698-023-00421-9 |
| _version_ | 1797425974314270720 |
|---|---|
| author | Yu Fu Marie Karanian Raul Perret Axel Camara François Le Loarer Myriam Jean-Denis Isabelle Hostein Audrey Michot Françoise Ducimetiere Antoine Giraud Jean-Baptiste Courreges Kevin Courtet Yech’an Laizet Etienne Bendjebbar Jean Ogier Du Terrail Benoit Schmauch Charles Maussion Jean-Yves Blay Antoine Italiano Jean-Michel Coindre |
| author_facet | Yu Fu Marie Karanian Raul Perret Axel Camara François Le Loarer Myriam Jean-Denis Isabelle Hostein Audrey Michot Françoise Ducimetiere Antoine Giraud Jean-Baptiste Courreges Kevin Courtet Yech’an Laizet Etienne Bendjebbar Jean Ogier Du Terrail Benoit Schmauch Charles Maussion Jean-Yves Blay Antoine Italiano Jean-Michel Coindre |
| author_sort | Yu Fu |
| collection | DOAJ |
| description | Abstract Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, the AFIP/Miettinen classification depends heavily on mitotic counts, which is laborious and sometimes inconsistent between pathologists. It has also been shown to be imperfect in stratifying patients. Molecular testing is costly and time-consuming, therefore, not systematically performed in all countries. New methods to improve risk and molecular predictions are hence crucial to improve the tailoring of adjuvant therapy. We have built deep learning (DL) models on digitized HES-stained whole slide images (WSI) to predict patients’ outcome and mutations. Models were trained with a cohort of 1233 GIST and validated on an independent cohort of 286 GIST. DL models yielded comparable results to the Miettinen classification for relapse-free-survival prediction in localized GIST without adjuvant Imatinib (C-index=0.83 in cross-validation and 0.72 for independent testing). DL splitted Miettinen intermediate risk GIST into high/low-risk groups (p value = 0.002 in the training set and p value = 0.29 in the testing set). DL models achieved an area under the receiver operating characteristic curve (AUC) of 0.81, 0.91, and 0.71 for predicting mutations in KIT, PDGFRA and wild type, respectively, in cross-validation and 0.76, 0.90, and 0.55 in independent testing. Notably, PDGFRA exon18 D842V mutation, which is resistant to Imatinib, was predicted with an AUC of 0.87 and 0.90 in cross-validation and independent testing, respectively. Additionally, novel histological criteria predictive of patients’ outcome and mutations were identified by reviewing the tiles selected by the models. As a proof of concept, our study showed the possibility of implementing DL with digitized WSI and may represent a reproducible way to improve tailoring therapy and precision medicine for patients with GIST. |
| first_indexed | 2024-03-09T08:23:42Z |
| format | Article |
| id | doaj.art-75a03086c3eb458f956500ff25fbf5f3 |
| institution | Directory Open Access Journal |
| issn | 2397-768X |
| language | English |
| last_indexed | 2024-03-09T08:23:42Z |
| publishDate | 2023-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Precision Oncology |
| spelling | doaj.art-75a03086c3eb458f956500ff25fbf5f32023-12-02T21:36:17ZengNature Portfolionpj Precision Oncology2397-768X2023-07-01711910.1038/s41698-023-00421-9Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumorYu Fu0Marie Karanian1Raul Perret2Axel Camara3François Le Loarer4Myriam Jean-Denis5Isabelle Hostein6Audrey Michot7Françoise Ducimetiere8Antoine Giraud9Jean-Baptiste Courreges10Kevin Courtet11Yech’an Laizet12Etienne Bendjebbar13Jean Ogier Du Terrail14Benoit Schmauch15Charles Maussion16Jean-Yves Blay17Antoine Italiano18Jean-Michel Coindre19Owkin, Inc.Cancer Research Center of Lyon, Centre Léon BérardDepartment of Biopathology, Institut BergoniéOwkin, Inc.Department of Biopathology, Institut BergoniéCancer Research Center of Lyon, Centre Léon BérardDepartment of Biopathology, Institut BergoniéDepartment of Surgical Oncology, Institut BergoniéCancer Research Center of Lyon, Centre Léon BérardClinical Research and Clinical Epidemiology Unit, Institut BergoniéClinical Research and Clinical Epidemiology Unit, Institut BergoniéDepartment of Biopathology, Institut BergoniéDepartment of Biopathology, Institut BergoniéOwkin, Inc.Owkin, Inc.Owkin, Inc.Owkin, Inc.Cancer Research Center of Lyon, Centre Léon BérardFaculty of Medicine, University of BordeauxDepartment of Biopathology, Institut BergoniéAbstract Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, the AFIP/Miettinen classification depends heavily on mitotic counts, which is laborious and sometimes inconsistent between pathologists. It has also been shown to be imperfect in stratifying patients. Molecular testing is costly and time-consuming, therefore, not systematically performed in all countries. New methods to improve risk and molecular predictions are hence crucial to improve the tailoring of adjuvant therapy. We have built deep learning (DL) models on digitized HES-stained whole slide images (WSI) to predict patients’ outcome and mutations. Models were trained with a cohort of 1233 GIST and validated on an independent cohort of 286 GIST. DL models yielded comparable results to the Miettinen classification for relapse-free-survival prediction in localized GIST without adjuvant Imatinib (C-index=0.83 in cross-validation and 0.72 for independent testing). DL splitted Miettinen intermediate risk GIST into high/low-risk groups (p value = 0.002 in the training set and p value = 0.29 in the testing set). DL models achieved an area under the receiver operating characteristic curve (AUC) of 0.81, 0.91, and 0.71 for predicting mutations in KIT, PDGFRA and wild type, respectively, in cross-validation and 0.76, 0.90, and 0.55 in independent testing. Notably, PDGFRA exon18 D842V mutation, which is resistant to Imatinib, was predicted with an AUC of 0.87 and 0.90 in cross-validation and independent testing, respectively. Additionally, novel histological criteria predictive of patients’ outcome and mutations were identified by reviewing the tiles selected by the models. As a proof of concept, our study showed the possibility of implementing DL with digitized WSI and may represent a reproducible way to improve tailoring therapy and precision medicine for patients with GIST.https://doi.org/10.1038/s41698-023-00421-9 |
| spellingShingle | Yu Fu Marie Karanian Raul Perret Axel Camara François Le Loarer Myriam Jean-Denis Isabelle Hostein Audrey Michot Françoise Ducimetiere Antoine Giraud Jean-Baptiste Courreges Kevin Courtet Yech’an Laizet Etienne Bendjebbar Jean Ogier Du Terrail Benoit Schmauch Charles Maussion Jean-Yves Blay Antoine Italiano Jean-Michel Coindre Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor npj Precision Oncology |
| title | Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| title_full | Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| title_fullStr | Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| title_full_unstemmed | Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| title_short | Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| title_sort | deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor |
| url | https://doi.org/10.1038/s41698-023-00421-9 |
| work_keys_str_mv | AT yufu deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT mariekaranian deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT raulperret deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT axelcamara deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT francoisleloarer deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT myriamjeandenis deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT isabellehostein deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT audreymichot deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT francoiseducimetiere deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT antoinegiraud deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT jeanbaptistecourreges deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT kevincourtet deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT yechanlaizet deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT etiennebendjebbar deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT jeanogierduterrail deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT benoitschmauch deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT charlesmaussion deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT jeanyvesblay deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT antoineitaliano deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor AT jeanmichelcoindre deeplearningpredictspatientsoutcomeandmutationsfromdigitizedhistologyslidesingastrointestinalstromaltumor |