Pitfalls in machine learning‐based assessment of tumor‐infiltrating lymphocytes in breast cancer: A report of the International Immuno‐Oncology Biomarker Working Group on Breast Cancer
Jeppe Thagaard, Glenn Broeckx, David B. Page, Chowdhury Arif Jahangir, Sara Verbandt, Zuzana Kos, Rajarsi Gupta, Reena Khiroya, Khalid AbdulJabbar, Gabriela Acosta Haab, Balázs Ács, Güray Aktürk, Jonas S. Almeida, Isabel Alvarado‐Cabrero, Mohamed Amgad, Farid Azmoudeh Ardalan, Sunil Badve, Nurkhairul Bariyah Baharun, Eva Balslev, Enrique Bellolio, Vydehi Bheemaraju, Kim RM Blenman, Luciana Botinelly Mendonça Fujimoto, Najat Bouchmaa, Octavio Burgues, Alexandros Hardas, Maggie C.U. Cheang, Francesco Ciompi, Lee Cooper, An Coosemans, Germán Corredor, Anders Bjorholm Dahl, Flávio Luis Dantas Portela, Frederik Deman, Sandra Demaria, Johan Doré Hansen, Sarah Dudgeon, Thomas Ebstrup, Mahmoud Elghazawy, Claudio Fernandez‐Martín, Stephen B. Fox, William M. Gallagher, Jennifer M. Giltnane, Sacha Gnjatic, Paula I. González-Ericsson, Anita Grigoriadis, Niels Halama, Matthew G Hanna, Aparna Harbhajanka, Steven N. Hart, Johan Hartman, Søren Hauberg, Stephen M. Hewitt, Akira I. Hida, Hugo M. Horlings, Zaheed Husain, Evangelos Hytopoulos, Sheeba Irshad, Emiel A. M. Janssen, Mohamed M. Kahila, Tatsuki R. Kataoka, Kosuke Kawaguchi, Kharidehal Durga, Andrey Khramtsov, Umay Kiraz, Pawan Kirtani, Liudmila L. Kodach, Konstanty Korski, Anikó Kovács, Anne‐Vibeke Lænkholm, Corinna Lang‐Schwarz, Denis Larsimont, Jochen K. Lennerz, Marvin Lerousseau, Xiaoxian Li, Amy Ly, Anant Madabhushi, Sai Maley, Vidya Manur Narasimhamurthy, Douglas K. Marks, Elizabeth S. McDonald, Ravi Mehrotra, Stefan Michiels, Fayyaz Minhas, Shachi Mittal, David A. Moore, Shamim Mushtaq, Nighat Hussain, Thomas Papathomas, Frédérique Penault‐Llorca, Rashindrie Perera, Christopher J. Pinard, Juan Carlos Pinto‐Cardenas, Giancarlo Pruneri, Lajos Pusztai, Arman Rahman, Nasir Rajpoot, Bernardo L. Rapoport, Tilman T. Rau, Jorge S. Reis‐Filho
Abstract
The clinical significance of the tumor-immune interaction in breast cancer is now established, and tumor-infiltrating lymphocytes (TILs) have emerged as predictive and prognostic biomarkers for patients with triple-negative (estrogen receptor, progesterone receptor, and HER2-negative) breast cancer and HER2-positive breast cancer. How computational assessments of TILs might complement manual TIL assessment in trial and daily practices is currently debated. Recent efforts to use machine learning (ML) to automatically evaluate TILs have shown promising results. We review state-of-the-art approaches and identify pitfalls and challenges of automated TIL evaluation by studying the root cause of ML discordances in comparison to manual TIL quantification. We categorize our findings into four main topics: (1) technical slide issues, (2) ML and image analysis aspects, (3) data challenges, and (4) validation issues. The main reason for discordant assessments is the inclusion of false-positive areas or cells identified by performance on certain tissue patterns or design choices in the computational implementation. To aid the adoption of ML for TIL assessment, we provide an in-depth discussion of ML and image analysis, including validation issues that need to be considered before reliable computational reporting of TILs can be incorporated into the trial and routine clinical management of patients with triple-negative breast cancer. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.