Related Experiment Video
Updated: Jun 25, 2026
09:53
Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020
Artificial Intelligence for Molecular Subtyping in Unresectable Gallbladder Cancer: A Proof-of-Concept Study for
Pankaj Gupta1,2, Chetan Madan2, Niharika Dutta1
1Department of Radiodiagnosis, Postgraduate Institute of Medical Education and Research, Chandigarh, 160012, India.
Journal of Clinical and Experimental Hepatology
|June 17, 2026
Summary
We developed an automated CT scan analysis to predict HER2 status in gallbladder cancer (GBC) without invasive biopsies. Deep learning models showed promising accuracy, offering a non-invasive approach for HER2-positive GBC detection.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Human epidermal growth factor receptor 2 (HER2) overexpression is a key therapeutic target in gallbladder cancer (GBC).
- Current HER2 detection methods for GBC require invasive tissue sampling.
- Non-invasive prediction of HER2 status is crucial for unresectable GBC management.
Purpose of the Study:
- To develop and validate a fully automated computed tomography (CT)-based framework for non-invasive HER2 status prediction in unresectable GBC.
- To compare the performance of clinical, radiomics, and deep learning (DL) models in predicting HER2 status.
Main Methods:
- A two-stage framework was developed: automated tumor detection/segmentation using Grounding DINO-MedSAM and HER2 classification.
- Classification models included clinical, radiomics, and DL approaches (Swin Transformer, DenseNet) with attention mechanisms.
- Performance was evaluated using Dice, IoU, sensitivity, specificity, AUROC, and F1 scores on a cohort of 213 GBC patients.
Main Results:
- The automated segmentation pipeline achieved a mean Dice score of 0.62 and IoU of 0.53.
- DL models outperformed clinical and radiomics models.
- The Swin Transformer DL model achieved an AUROC of 0.792 and an F1-score of 0.769 on the test set.
- No significant performance difference was observed between automated and ground truth segmentations.
Conclusions:
- A fully automated CT-based DL pipeline can predict HER2 status in unresectable GBC.
- This non-invasive approach shows potential for improving GBC treatment strategies.
- Further validation in larger, multi-institutional cohorts is recommended.
