RT Journal Article
SR Electronic
T1 Spatial analysis of tumor infiltrating lymphocytes based on deep learning using histopathology image to predict progression-free survival in colorectal cancer
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2021.04.24.441275
DO 10.1101/2021.04.24.441275
A1 Hongming Xu
A1 Yoon Jin Cha
A1 Jean R. Clemenceau
A1 Jinhwan Choi
A1 Sung Hak Lee
A1 Jeonghyun Kang
A1 Tae Hyun Hwang
YR 2021
UL http://biorxiv.org/content/early/2021/04/27/2021.04.24.441275.abstract
AB Purpose This study aimed to explore the prognostic impact of spatial distribution of tumor infiltrating lymphocytes (TILs) quantified by deep learning (DL) approaches based on digitalized whole slide images stained with hematoxylin and eosin in patients with colorectal cancer (CRC).Methods The prognostic impact of spatial distributions of TILs in patients with CRC was explored in the Yonsei cohort (n=180) and validated in the TCGA cohort (n=268). Concurrently, two experienced pathologists manually measured TILs at the most invasive margin as 0-3 by the Klintrup-Mäkinen (KM) grading method and compared to DL approaches. Interobserver agreement for TILs was measured using Cohen’s kappa coefficient.Results On multivariate analysis of spatial TILs features derived by DL approaches and clinicopathological variables including tumor stage, Microsatellite instability, and KRAS mutations, TILs densities within 200 μm of the invasive margin (f_im200) was remained as the most significant prognostic factor for progression-free survival (PFS) (HR 0.004 [95% CI, 0.0001-0.1502], p=.002) in the Yonsei cohort. On multivariate analysis using the TCGA dataset, f_im200 retained prognostic significance for PFS (HR 0.031, [95% CI 0.001-0.645], p=.024). Interobserver agreement of manual KM grading based on Cohen’s kappa coefficient was insignificant in the Yonsei (κ=.109) and the TCGA (κ=.121), respectively. The survival analysis based on KM grading showed statistically significant different PFS from the TCGA cohort, but not the Yonsei cohort.Conclusions Automatic quantification of TILs at the invasive margin based on DL approaches showed a prognostic utility to predict PFS, and could provide robust and reproducible TILs density measurement in patients with CRC.Data and Code Availability Source code and data used for this study is available at the following link: https://github.com/hwanglab/TILs_AnalysisCompeting Interest StatementTHH received consulting fees from AITRICS. THH received research funding from AITRICS through the insititute. THH is co-founder of KURE.AI.