Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer

IntroductionProgrammed cell death ligand-1 (PD-L1) combined positive score (CPS) evaluation plays a pivotal role in predicting immunotherapy efficacy for gastric cancer. However, manual CPS assessment suffers from significant inter-observer variability among pathologists, leading to clinical inconsi...

Full description

Saved in:
Bibliographic Details
Main Authors: Ting Han, Meng Zhuo, Ziyu Song, Peilin Chen, Shiting Chen, Wei Zhang, Yuanyuan Zhou, Hong Li, Dadong Zhang, Xiaolin Lin, Zebing Liu, Xiuying Xiao
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-08-01
Series:Frontiers in Immunology
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fimmu.2025.1614099/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849395662549942272
author Ting Han
Meng Zhuo
Ziyu Song
Peilin Chen
Shiting Chen
Wei Zhang
Yuanyuan Zhou
Hong Li
Dadong Zhang
Xiaolin Lin
Zebing Liu
Xiuying Xiao
author_facet Ting Han
Meng Zhuo
Ziyu Song
Peilin Chen
Shiting Chen
Wei Zhang
Yuanyuan Zhou
Hong Li
Dadong Zhang
Xiaolin Lin
Zebing Liu
Xiuying Xiao
author_sort Ting Han
collection DOAJ
description IntroductionProgrammed cell death ligand-1 (PD-L1) combined positive score (CPS) evaluation plays a pivotal role in predicting immunotherapy efficacy for gastric cancer. However, manual CPS assessment suffers from significant inter-observer variability among pathologists, leading to clinical inconsistencies. To address this limitation, we developed a deep learning-based artificial intelligence (AI) system that automates PD-L1 CPS quantification for patients with gastric cancer (GC) using whole slide images (WSIs).MethodsWe developed a deep learning-based artificial intelligence (AI) system that automates PD-L1 CPS quantification for patients with gastric cancer (GC) using whole slide images (WSIs). Our pipeline firstly employs a dual-network architecture for tumor region detection: MobileNet for patch-level classification and U-Net for pixel-level segmentation. Followed by a YOLO-based cell detection model to compute PD-L1 expression on different cells for CPS calculation. A total of 308 GC WSIs were included, including 210 in the internal cohort and 98 in the external cohort. Within the internal cohort, 100 WSIs were utilized for the model development, while the remaining 110 WSIs served as an internal testing set for comparative analysis between AI-derived CPS values and pathologist-derived reference standards.ResultsThe AI-derived CPS demonstrated strong concordance with expert pathologists’ consensus in internal cohort (Cohen’s kappa = 0.782). Furthermore, the AI-based CPS prediction pipeline was evaluated for its performance in the external cohort, and showed robust performance (Cohen’s kappa = 0.737).DiscussionOur system provides a standardized decision-support tool for immunotherapy stratification in GC management, demonstrating potential to improve CPS assessment reproducibility.
format Article
id doaj-art-9d63ed820c7c409aa43b94ffa9b33bc7
institution Kabale University
issn 1664-3224
language English
publishDate 2025-08-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Immunology
spelling doaj-art-9d63ed820c7c409aa43b94ffa9b33bc72025-08-20T03:39:32ZengFrontiers Media S.A.Frontiers in Immunology1664-32242025-08-011610.3389/fimmu.2025.16140991614099Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancerTing Han0Meng Zhuo1Ziyu Song2Peilin Chen3Shiting Chen4Wei Zhang5Yuanyuan Zhou6Hong Li7Dadong Zhang8Xiaolin Lin9Zebing Liu10Xiuying Xiao11Department of Oncology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaDepartment of Oncology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaDepartment of Pathology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaProduct Development Department, SODA Data Technology Inc., Shanghai, ChinaProduct Development Department, SODA Data Technology Inc., Shanghai, ChinaDepartment of Clinical and Translational Research, 3D Medicines Inc., Shanghai, ChinaSchool of Pharmacy, East China University of Science and Technology, Shanghai, ChinaState Key Laboratory of Systems Medicine for Cancer, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, ChinaDepartment of Clinical and Translational Research, 3D Medicines Inc., Shanghai, ChinaDepartment of Oncology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaDepartment of Pathology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaDepartment of Oncology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, ChinaIntroductionProgrammed cell death ligand-1 (PD-L1) combined positive score (CPS) evaluation plays a pivotal role in predicting immunotherapy efficacy for gastric cancer. However, manual CPS assessment suffers from significant inter-observer variability among pathologists, leading to clinical inconsistencies. To address this limitation, we developed a deep learning-based artificial intelligence (AI) system that automates PD-L1 CPS quantification for patients with gastric cancer (GC) using whole slide images (WSIs).MethodsWe developed a deep learning-based artificial intelligence (AI) system that automates PD-L1 CPS quantification for patients with gastric cancer (GC) using whole slide images (WSIs). Our pipeline firstly employs a dual-network architecture for tumor region detection: MobileNet for patch-level classification and U-Net for pixel-level segmentation. Followed by a YOLO-based cell detection model to compute PD-L1 expression on different cells for CPS calculation. A total of 308 GC WSIs were included, including 210 in the internal cohort and 98 in the external cohort. Within the internal cohort, 100 WSIs were utilized for the model development, while the remaining 110 WSIs served as an internal testing set for comparative analysis between AI-derived CPS values and pathologist-derived reference standards.ResultsThe AI-derived CPS demonstrated strong concordance with expert pathologists’ consensus in internal cohort (Cohen’s kappa = 0.782). Furthermore, the AI-based CPS prediction pipeline was evaluated for its performance in the external cohort, and showed robust performance (Cohen’s kappa = 0.737).DiscussionOur system provides a standardized decision-support tool for immunotherapy stratification in GC management, demonstrating potential to improve CPS assessment reproducibility.https://www.frontiersin.org/articles/10.3389/fimmu.2025.1614099/fullPD-L1CPSgastric cancerautomated scoringartificial intelligence
spellingShingle Ting Han
Meng Zhuo
Ziyu Song
Peilin Chen
Shiting Chen
Wei Zhang
Yuanyuan Zhou
Hong Li
Dadong Zhang
Xiaolin Lin
Zebing Liu
Xiuying Xiao
Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
Frontiers in Immunology
PD-L1
CPS
gastric cancer
automated scoring
artificial intelligence
title Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
title_full Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
title_fullStr Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
title_full_unstemmed Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
title_short Automated interpretation of PD-L1 CPS based on multi-AI models integration strategy in gastric cancer
title_sort automated interpretation of pd l1 cps based on multi ai models integration strategy in gastric cancer
topic PD-L1
CPS
gastric cancer
automated scoring
artificial intelligence
url https://www.frontiersin.org/articles/10.3389/fimmu.2025.1614099/full
work_keys_str_mv AT tinghan automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT mengzhuo automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT ziyusong automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT peilinchen automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT shitingchen automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT weizhang automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT yuanyuanzhou automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT hongli automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT dadongzhang automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT xiaolinlin automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT zebingliu automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer
AT xiuyingxiao automatedinterpretationofpdl1cpsbasedonmultiaimodelsintegrationstrategyingastriccancer