TY - JOUR
T1 - End-to-end prognostication in colorectal cancer by deep learning
T2 - a retrospective, multicentre study
AU - Jiang, Xiaofeng
AU - Hoffmeister, Michael
AU - Brenner, Hermann
AU - Muti, Hannah Sophie
AU - Yuan, Tanwei
AU - Foersch, Sebastian
AU - West, Nicholas P.
AU - Brobeil, Alexander
AU - Jonnagaddala, Jitendra
AU - Hawkins, Nicholas
AU - Ward, Robyn L.
AU - Brinker, Titus J.
AU - Saldanha, Oliver Lester
AU - Ke, Jia
AU - Müller, Wolfram
AU - Grabsch, Heike I.
AU - Quirke, Philip
AU - Truhn, Daniel
AU - Kather, Jakob Nikolas
N1 - Funding Information:
We extend our gratitude to the tissue bank of the National Center for Tumor Diseases at the Institute of Pathology at University Hospital Heidelberg, Germany for providing access to the biobank data. We also acknowledge the assistance of the SREDH Consortium's ( www.sredhconsortium.org ) Translational Cancer Bioinformatics working group in obtaining access to the Molecular and Cellular Oncology colorectal cancer dataset. JNK is supported by the German Federal Ministry of Health (Deep Liver, ZMVI1-2520DAT111), the Max-Eder-Programme of German Cancer Aid (grant 70113864), the German Federal Ministry of Education and Research (PEARL, 01KD2104C; Camino, 01EO2101; SWAG, 01KD2215A; Transform Liver, 031L0312A and Tangerine, 01KT2302 through ERA-NET Transcan), the German Academic Exchange Service (SECAI, 57616814), the German Federal Joint Committee (Transplant.KI, 01VSF21048) and the EU's Horizon Europe innovation programme (Odelia, 101057091; Genial, 101096312). JNK, PQ, and NPW are supported by the National Institute for Health and Care Research (NIHR; NIHR213331) Leeds Biomedical Research Centre. XJ is supported by the programme of the China Scholarships Council (202106380048). The Darmkrebs: Chancen der Verhütung durch Screening study (HB and MH) was supported by the German Research Council (BR 1704/6-1, BR 1704/6-3, BR 1704/6-4, CH 117/1-1, HO 5117/2-1, HO 5117/2-2, HE 5998/2-1, HE 5998/2-2, KL 2354/3-1, KL 2354/3-2, RO 2270/8-1, RO 2270/8-2, BR 1704/17-1, and BR 1704/17-2), the Interdisciplinary Research Program of the National Centre for Tumour Diseases (Germany), and the German Federal Ministry of Education and Research (01KH0404, 01ER0814, 01ER0815, 01ER1505A, and 01ER1505B). PQ and NPW are supported by Yorkshire Cancer Research Programme grants L386 (QUASAR series) and L394 (YCR BCIP series). PQ is a National Institute of Health Research senior investigator. JJ was funded by the Australian National Health and Medical Research Council (GNT1192469) and was also supported by Google through the 2022 research innovator and cloud research credits programme (GCP19980904) and the Research Technology Services at University of New South Wales Sydney, and NVIDIA Academic Hardware grant programmes.
Funding Information:
We extend our gratitude to the tissue bank of the National Center for Tumor Diseases at the Institute of Pathology at University Hospital Heidelberg, Germany for providing access to the biobank data. We also acknowledge the assistance of the SREDH Consortium's (www.sredhconsortium.org) Translational Cancer Bioinformatics working group in obtaining access to the Molecular and Cellular Oncology colorectal cancer dataset. JNK is supported by the German Federal Ministry of Health (Deep Liver, ZMVI1-2520DAT111), the Max-Eder-Programme of German Cancer Aid (grant 70113864), the German Federal Ministry of Education and Research (PEARL, 01KD2104C; Camino, 01EO2101; SWAG, 01KD2215A; Transform Liver, 031L0312A and Tangerine, 01KT2302 through ERA-NET Transcan), the German Academic Exchange Service (SECAI, 57616814), the German Federal Joint Committee (Transplant.KI, 01VSF21048) and the EU's Horizon Europe innovation programme (Odelia, 101057091; Genial, 101096312). JNK, PQ, and NPW are supported by the National Institute for Health and Care Research (NIHR; NIHR213331) Leeds Biomedical Research Centre. XJ is supported by the programme of the China Scholarships Council (202106380048). The Darmkrebs: Chancen der Verhütung durch Screening study (HB and MH) was supported by the German Research Council (BR 1704/6-1, BR 1704/6-3, BR 1704/6-4, CH 117/1-1, HO 5117/2-1, HO 5117/2-2, HE 5998/2-1, HE 5998/2-2, KL 2354/3-1, KL 2354/3-2, RO 2270/8-1, RO 2270/8-2, BR 1704/17-1, and BR 1704/17-2), the Interdisciplinary Research Program of the National Centre for Tumour Diseases (Germany), and the German Federal Ministry of Education and Research (01KH0404, 01ER0814, 01ER0815, 01ER1505A, and 01ER1505B). PQ and NPW are supported by Yorkshire Cancer Research Programme grants L386 (QUASAR series) and L394 (YCR BCIP series). PQ is a National Institute of Health Research senior investigator. JJ was funded by the Australian National Health and Medical Research Council (GNT1192469) and was also supported by Google through the 2022 research innovator and cloud research credits programme (GCP19980904) and the Research Technology Services at University of New South Wales Sydney, and NVIDIA Academic Hardware grant programmes.
Publisher Copyright:
© 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license
PY - 2024/1/1
Y1 - 2024/1/1
N2 - Background Precise prognosis prediction in patients with colorectal cancer (ie, forecasting survival) is pivotal for individualised treatment and care. Histopathological tissue slides of colorectal cancer specimens contain rich prognostically relevant information. However, existing studies do not have multicentre external validation with real-world sample processing protocols, and algorithms are not yet widely used in clinical routine.Methods In this retrospective, multicentre study, we collected tissue samples from four groups of patients with resected colorectal cancer from Australia, Germany, and the USA. We developed and externally validated a deep learning-based prognostic-stratification system for automatic prediction of overall and cancer-specific survival in patients with resected colorectal cancer. We used the model-predicted risk scores to stratify patients into different risk groups and compared survival outcomes between these groups. Additionally, we evaluated the prognostic value of these risk groups after adjusting for established prognostic variables.Findings We trained and validated our model on a total of 4428 patients. We found that patients could be divided into high-risk and low-risk groups on the basis of the deep learning-based risk score. On the internal test set, the group with a high-risk score had a worse prognosis than the group with a low-risk score, as reflected by a hazard ratio (HR) of 4 center dot 50 (95% CI 3 center dot 33-6 center dot 09) for overall survival and 8 center dot 35 (5 center dot 06-13 center dot 78) for disease-specific survival (DSS). We found consistent performance across three large external test sets. In a test set of 1395 patients, the high-risk group had a lower DSS than the low-risk group, with an HR of 3 center dot 08 (2 center dot 44-3 center dot 89). In two additional test sets, the HRs for DSS were 2 center dot 23 (1 center dot 23-4 center dot 04) and 3 center dot 07 (1 center dot 78-5 center dot 3). We showed that the prognostic value of the deep learning-based risk score is independent of established clinical risk factors.Interpretation Our findings indicate that attention-based self-supervised deep learning can robustly offer a prognosis on clinical outcomes in patients with colorectal cancer, generalising across different populations and serving as a potentially new prognostic tool in clinical decision making for colorectal cancer management. We release all source codes and trained models under an open-source licence, allowing other researchers to reuse and build upon our work. Funding The German Federal Ministry of Health, the Max-Eder-Programme of German Cancer Aid, the German Federal Ministry of Education and Research, the German Academic Exchange Service, and the EU.Copyright (c) 2023 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.
AB - Background Precise prognosis prediction in patients with colorectal cancer (ie, forecasting survival) is pivotal for individualised treatment and care. Histopathological tissue slides of colorectal cancer specimens contain rich prognostically relevant information. However, existing studies do not have multicentre external validation with real-world sample processing protocols, and algorithms are not yet widely used in clinical routine.Methods In this retrospective, multicentre study, we collected tissue samples from four groups of patients with resected colorectal cancer from Australia, Germany, and the USA. We developed and externally validated a deep learning-based prognostic-stratification system for automatic prediction of overall and cancer-specific survival in patients with resected colorectal cancer. We used the model-predicted risk scores to stratify patients into different risk groups and compared survival outcomes between these groups. Additionally, we evaluated the prognostic value of these risk groups after adjusting for established prognostic variables.Findings We trained and validated our model on a total of 4428 patients. We found that patients could be divided into high-risk and low-risk groups on the basis of the deep learning-based risk score. On the internal test set, the group with a high-risk score had a worse prognosis than the group with a low-risk score, as reflected by a hazard ratio (HR) of 4 center dot 50 (95% CI 3 center dot 33-6 center dot 09) for overall survival and 8 center dot 35 (5 center dot 06-13 center dot 78) for disease-specific survival (DSS). We found consistent performance across three large external test sets. In a test set of 1395 patients, the high-risk group had a lower DSS than the low-risk group, with an HR of 3 center dot 08 (2 center dot 44-3 center dot 89). In two additional test sets, the HRs for DSS were 2 center dot 23 (1 center dot 23-4 center dot 04) and 3 center dot 07 (1 center dot 78-5 center dot 3). We showed that the prognostic value of the deep learning-based risk score is independent of established clinical risk factors.Interpretation Our findings indicate that attention-based self-supervised deep learning can robustly offer a prognosis on clinical outcomes in patients with colorectal cancer, generalising across different populations and serving as a potentially new prognostic tool in clinical decision making for colorectal cancer management. We release all source codes and trained models under an open-source licence, allowing other researchers to reuse and build upon our work. Funding The German Federal Ministry of Health, the Max-Eder-Programme of German Cancer Aid, the German Federal Ministry of Education and Research, the German Academic Exchange Service, and the EU.Copyright (c) 2023 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.
KW - PREDICTION
KW - DIAGNOSIS
U2 - 10.1016/S2589-7500(23)00208-X
DO - 10.1016/S2589-7500(23)00208-X
M3 - Article
SN - 2589-7500
VL - 6
SP - e33-e43
JO - The Lancet Digital Health
JF - The Lancet Digital Health
IS - 1
ER -