Amelie Echle1, Heike Irmgard Grabsch2, Philip Quirke3, Piet A van den Brandt4, Nicholas P West3, Gordon G A Hutchins3, Lara R Heij5, Xiuxiang Tan5, Susan D Richman3, Jeremias Krause1, Elizabeth Alwers6, Josien Jenniskens4, Kelly Offermans4, Richard Gray7, Hermann Brenner8, Jenny Chang-Claude9, Christian Trautwein1, Alexander T Pearson10, Peter Boor11, Tom Luedde12, Nadine Therese Gaisa11, Michael Hoffmeister6, Jakob Nikolas Kather13. 1. Department of Medicine III, University Hospital RWTH Aachen, Aachen, Germany. 2. Department of Pathology, GROW School for Oncology and Developmental Biology, Maastricht University Medical Center+, Maastricht, The Netherlands; Pathology and Data Analytics, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, United Kingdom. 3. Pathology and Data Analytics, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, United Kingdom. 4. Department of Epidemiology, Maastricht University Medical Center+, Maastricht, The Netherlands. 5. Visceral and Transplant Surgery, University Hospital Rheinisch-Westfälische Technische Hochschule Aachen, Aachen, Germany; NUTRIM School of Nutrition and Translational Research in Metabolism, Maastricht University, Maastricht, the Netherlands; Institute of Pathology, University Hospital RWTH Aachen, Aachen, Germany. 6. Division of Clinical Epidemiology and Aging Research, German Cancer Research Center, Heidelberg, Germany. 7. Clinical Trial Service Unit, University of Oxford, Oxford, United Kingdom. 8. Division of Clinical Epidemiology and Aging Research, German Cancer Research Center, Heidelberg, Germany; Division of Preventive Oncology, German Cancer Research Center and National Center for Tumor Diseases, Heidelberg, Germany; German Cancer Consortium, German Cancer Research Center, Heidelberg, Germany. 9. Division of Cancer Epidemiology, German Cancer Research Center, Heidelberg, Germany; Cancer Epidemiology Group, University Cancer Center Hamburg, University Medical Center Hamburg-Eppendorf, Hamburg, Germany. 10. Section of Hematology/Oncology, Department of Medicine, University of Chicago, Chicago, Illinois. 11. Institute of Pathology, University Hospital RWTH Aachen, Aachen, Germany. 12. Department of Medicine III, University Hospital RWTH Aachen, Aachen, Germany; Division of Gastroenterology, Hepatology, and Hepatobiliary Oncology, Aachen, Germany. 13. Department of Medicine III, University Hospital RWTH Aachen, Aachen, Germany; Pathology and Data Analytics, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, United Kingdom; German Cancer Consortium, German Cancer Research Center, Heidelberg, Germany; Medical Oncology, National Center for Tumor Diseases, University Hospital Heidelberg, Heidelberg, Germany. Electronic address: jkather@ukaachen.de.
Abstract
BACKGROUND & AIMS: Microsatellite instability (MSI) and mismatch-repair deficiency (dMMR) in colorectal tumors are used to select treatment for patients. Deep learning can detect MSI and dMMR in tumor samples on routine histology slides faster and less expensively than molecular assays. However, clinical application of this technology requires high performance and multisite validation, which have not yet been performed. METHODS: We collected H&E-stained slides and findings from molecular analyses for MSI and dMMR from 8836 colorectal tumors (of all stages) included in the MSIDETECT consortium study, from Germany, the Netherlands, the United Kingdom, and the United States. Specimens with dMMR were identified by immunohistochemistry analyses of tissue microarrays for loss of MLH1, MSH2, MSH6, and/or PMS2. Specimens with MSI were identified by genetic analyses. We trained a deep-learning detector to identify samples with MSI from these slides; performance was assessed by cross-validation (N = 6406 specimens) and validated in an external cohort (n = 771 specimens). Prespecified endpoints were area under the receiver operating characteristic (AUROC) curve and area under the precision-recall curve (AUPRC). RESULTS: The deep-learning detector identified specimens with dMMR or MSI with a mean AUROC curve of 0.92 (lower bound, 0.91; upper bound, 0.93) and an AUPRC of 0.63 (range, 0.59-0.65), or 67% specificity and 95% sensitivity, in the cross-validation development cohort. In the validation cohort, the classifier identified samples with dMMR with an AUROC of 0.95 (range, 0.92-0.96) without image preprocessing and an AUROC of 0.96 (range, 0.93-0.98) after color normalization. CONCLUSIONS: We developed a deep-learning system that detects colorectal cancer specimens with dMMR or MSI using H&E-stained slides; it detected tissues with dMMR with an AUROC of 0.96 in a large, international validation cohort. This system might be used for high-throughput, low-cost evaluation of colorectal tissue specimens.
BACKGROUND & AIMS:Microsatellite instability (MSI) and mismatch-repair deficiency (dMMR) in colorectal tumors are used to select treatment for patients. Deep learning can detect MSI and dMMR in tumor samples on routine histology slides faster and less expensively than molecular assays. However, clinical application of this technology requires high performance and multisite validation, which have not yet been performed. METHODS: We collected H&E-stained slides and findings from molecular analyses for MSI and dMMR from 8836 colorectal tumors (of all stages) included in the MSIDETECT consortium study, from Germany, the Netherlands, the United Kingdom, and the United States. Specimens with dMMR were identified by immunohistochemistry analyses of tissue microarrays for loss of MLH1, MSH2, MSH6, and/or PMS2. Specimens with MSI were identified by genetic analyses. We trained a deep-learning detector to identify samples with MSI from these slides; performance was assessed by cross-validation (N = 6406 specimens) and validated in an external cohort (n = 771 specimens). Prespecified endpoints were area under the receiver operating characteristic (AUROC) curve and area under the precision-recall curve (AUPRC). RESULTS: The deep-learning detector identified specimens with dMMR or MSI with a mean AUROC curve of 0.92 (lower bound, 0.91; upper bound, 0.93) and an AUPRC of 0.63 (range, 0.59-0.65), or 67% specificity and 95% sensitivity, in the cross-validation development cohort. In the validation cohort, the classifier identified samples with dMMR with an AUROC of 0.95 (range, 0.92-0.96) without image preprocessing and an AUROC of 0.96 (range, 0.93-0.98) after color normalization. CONCLUSIONS: We developed a deep-learning system that detects colorectal cancer specimens with dMMR or MSI using H&E-stained slides; it detected tissues with dMMR with an AUROC of 0.96 in a large, international validation cohort. This system might be used for high-throughput, low-cost evaluation of colorectal tissue specimens.
Authors: Yang Liu; Nilay S Sethi; Toshinori Hinoue; Barbara G Schneider; Andrew D Cherniack; Francisco Sanchez-Vega; Jose A Seoane; Farshad Farshidfar; Reanne Bowlby; Mirazul Islam; Jaegil Kim; Walid Chatila; Rehan Akbani; Rupa S Kanchi; Charles S Rabkin; Joseph E Willis; Kenneth K Wang; Shannon J McCall; Lopa Mishra; Akinyemi I Ojesina; Susan Bullman; Chandra Sekhar Pedamallu; Alexander J Lazar; Ryo Sakai; Vésteinn Thorsson; Adam J Bass; Peter W Laird Journal: Cancer Cell Date: 2018-04-02 Impact factor: 31.743
Authors: Tristan Snowsill; Helen Coelho; Nicola Huxley; Tracey Jones-Hughes; Simon Briscoe; Ian M Frayling; Chris Hyde Journal: Health Technol Assess Date: 2017-09 Impact factor: 4.014
Authors: Richard Gray; Jennifer Barnwell; Christopher McConkey; Robert K Hills; Norman S Williams; David J Kerr Journal: Lancet Date: 2007-12-15 Impact factor: 79.321
Authors: L De Smedt; J Lemahieu; S Palmans; O Govaere; T Tousseyn; E Van Cutsem; H Prenen; S Tejpar; M Spaepen; G Matthijs; C Decaestecker; X Moles Lopez; P Demetter; I Salmon; X Sagaert Journal: Br J Cancer Date: 2015-06-11 Impact factor: 7.640
Authors: Peter Bankhead; Maurice B Loughrey; José A Fernández; Yvonne Dombrowski; Darragh G McArt; Philip D Dunne; Stephen McQuaid; Ronan T Gray; Liam J Murray; Helen G Coleman; Jacqueline A James; Manuel Salto-Tellez; Peter W Hamilton Journal: Sci Rep Date: 2017-12-04 Impact factor: 4.379
Authors: Yoon-Jung Kang; James Killen; Michael Caruana; Kate Simms; Natalie Taylor; Ian M Frayling; Tristan Snowsill; Nicola Huxley; Veerle Mh Coupe; Suzanne Hughes; Victoria Freeman; Alex Boussioutas; Alison H Trainer; Robyn L Ward; Gillian Mitchell; Finlay A Macrae; Karen Canfell Journal: Med J Aust Date: 2019-10-08 Impact factor: 7.738
Authors: Oliver Lester Saldanha; Philip Quirke; Nicholas P West; Jacqueline A James; Maurice B Loughrey; Heike I Grabsch; Manuel Salto-Tellez; Elizabeth Alwers; Didem Cifci; Narmin Ghaffari Laleh; Tobias Seibel; Richard Gray; Gordon G A Hutchins; Hermann Brenner; Marko van Treeck; Tanwei Yuan; Titus J Brinker; Jenny Chang-Claude; Firas Khader; Andreas Schuppert; Tom Luedde; Christian Trautwein; Hannah Sophie Muti; Sebastian Foersch; Michael Hoffmeister; Daniel Truhn; Jakob Nikolas Kather Journal: Nat Med Date: 2022-04-25 Impact factor: 87.241
Authors: Marina Zimmermann; Martin Klaus; Milagros N Wong; Ann-Katrin Thebille; Lukas Gernhold; Christoph Kuppe; Maurice Halder; Jennifer Kranz; Nicola Wanner; Fabian Braun; Sonia Wulf; Thorsten Wiech; Ulf Panzer; Christian F Krebs; Elion Hoxha; Rafael Kramann; Tobias B Huber; Stefan Bonn; Victor G Puelles Journal: JCI Insight Date: 2021-04-08
Authors: Frederick M Howard; James Dolezal; Sara Kochanny; Jefree Schulte; Heather Chen; Lara Heij; Dezheng Huo; Rita Nanda; Olufunmilayo I Olopade; Jakob N Kather; Nicole Cipriani; Robert L Grossman; Alexander T Pearson Journal: Nat Commun Date: 2021-07-20 Impact factor: 14.919