Related Experiment Video
Updated: Jun 1, 2026
Industrialized, Artificial Intelligence-guided Laser Microdissection for Microscaled Proteomic Analysis of the Tumor Microenvironment
Published on: June 3, 2022
Deep learning based histomorphological phenotyping and prognostic stratification for combined SCLC and LCNEC
Lin Yang1,2, Ruyu Sheng3, Zijian Yang3
1Department of Pathology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, P. R. China. yanglin@cicams.ac.cn.
A new AI tool, GTBIS, accurately predicts outcomes for combined small and large cell neuroendocrine lung carcinoma (cSCLC-LCNEC) using pathology images. It identifies distinct patient subgroups with significantly different survival rates, aiding personalized treatment strategies.
Area of Science:
- Oncology
- Pathology
- Artificial Intelligence
Background:
- Prognostic stratification for combined small and large cell neuroendocrine lung carcinoma (cSCLC-LCNEC) is difficult.
- Accurate phenotyping from histopathology images is crucial for predicting patient outcomes.
Purpose of the Study:
- To introduce GTBIS, an interpretable deep learning model for histomorphological phenotyping and prognostic stratification of cSCLC-LCNEC.
- To evaluate GTBIS's performance in differentiating SCLC from LCNEC and stratifying cSCLC-LCNEC patients.
Main Methods:
- Development and application of GTBIS, a deep learning model, on pathology images from multicenter cohorts (n=670).
- Validation of GTBIS's prognostic capability in cSCLC-LCNEC patients treated with chemoradiotherapy.
- Multivariable analysis and multimodal interpretability analyses to confirm prognostic significance and biological correlates.
Main Results:
- GTBIS accurately differentiated SCLC from LCNEC.
- GTBIS stratified cSCLC-LCNEC patients into favorable-prognosis (SCLC-like) and poor-prognosis (LCNEC-like) subgroups.
- Favorable-prognosis subgroup showed significantly better 5-year overall survival (100% vs. 39.5%) and disease-free survival (87.5% vs. 36.0%).
- GTBIS classification was an independent prognostic factor.
Conclusions:
- GTBIS demonstrates potential as a histology-based tool for prognostic stratification in cSCLC-LCNEC.
- The model's interpretability links phenotypes to distinct biological pathways (proliferation, EMT, hypoxia, metabolism).
- GTBIS may facilitate more personalized management strategies for cSCLC-LCNEC patients.
