Contrastive multimodal deep learning for survival prediction in grade 2/3 gliomas

Key Points

• The study aims to improve survival prediction in grade 2/3 gliomas using a contrastive multimodal learning approach.
• Utilized contrastive multimodal learning techniques
• Integrated histopathology, genomics, and clinical data
• Developed an annotation-free model for risk stratification
• Demonstrated enhanced survival prediction accuracy
• Enabled early risk stratification with routine data
• Showed potential for informing personalized treatment and clinical trial stratification