MD Anderson's Path-IO machine learning platform accurately predicts immunotherapy responses in metastatic non-small cell lung cancer, surpassing current biomarker standards.
Key Details
- 1Path-IO uses routine pathological slides to identify features predicting immunotherapy response in NSCLC patients.
- 2In validation on over 1,000 patients across multiple institutions and countries, Path-IO significantly outperformed PD-L1 testing, the current standard biomarker.
- 3The AI model stratified patients into high- and low-risk groups, with high-risk patients experiencing double the probability of death or progression.
- 4Unlike 'black box' models, Path-IO leverages established tissue features for explainable, clinically translatable decision-making.
- 5Researchers are integrating radiomics and clinical data to further improve Path-IO's prediction accuracy and plan to include CT, genomics, and other multimodal data in the future.
- 6Prospective clinical validation is the next planned step, and the model may evolve into a digital twin framework for comprehensive clinical decision support.
Why It Matters
Path-IO's superior accuracy and explainability provide a pathway for more personalized and effective treatment decisions in oncology. The integration with radiomics and clinical imaging data signals a promising future for multimodal AI in precision medicine, especially for radiology and pathology collaborations.
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.