A deep learning pathomics platform accurately predicts immunotherapy response in metastatic NSCLC using routine pathology slides.
Key Details
- 1The Path-IO model uses AI to analyze digital pathology slides from NSCLC patients.
- 2Tested on 797 patients (MD Anderson) and validated on 280 international patients (including Lung-MAP S1400I trial).
- 3The model stratified patients into high- and low-risk groups, doubling risk in the high-risk cohort.
- 4Path-IO outperformed PD-L1 biomarker: Path-IO C-indices up to 0.69 (OS) and 0.65 (PFS); PD-L1 indices as low as 0.50-0.58.
- 5Integrating radiomics and clinical data improved prediction performance further (C-index for OS to 0.75, PFS to 0.70).
- 6Study validated across real-world and phase III trial cohorts, though retrospective; prospective validation is the next step.
Why It Matters
This work demonstrates that advanced AI analysis of digital pathology can outperform traditional biomarkers in predicting immunotherapy outcomes. If validated further, such tools could help personalize treatment for cancer patients, offering a significant advance in precision oncology and imaging AI.
Source
EurekAlert
Related News
•EurekAlert
AI-Powered OCT Enables Rapid 'Optical Biopsy' for Early Endometrial Cancer Detection
A team at Washington University has developed a catheter-based 3D OCT system with AI to quickly and noninvasively detect early endometrial cancers.
•EurekAlert
AI Clinical Reasoning in Diagnostics and Digital Fatigue in Healthcare
Recent JMIR features explore large language models in clinical diagnostics and digital fatigue among healthcare professionals.
•EurekAlert
KAIST, MIT, Microsoft Develop Efficient AI Image Upsampling for Robotics
KAIST, MIT, and Microsoft have created 'Upsample Anything,' a training-free AI method to restore high-resolution visual data from compressed images with up to 16x improved GPU memory efficiency.