MIF: Multi-Shot Interactive Fusion Model for Cancer Survival Prediction Using Pathological Image and Genomic Data.

Area of Science:

• Oncology
• Bioinformatics
• Medical Imaging

Background:

• Accurate cancer survival prediction is vital for treatment planning and patient outcomes.
• Multimodal fusion methods using pathological and genomic data show promise but face challenges in complex interaction modeling and computational efficiency.

Purpose of the Study:

• To propose an innovative multi-shot interactive fusion method (MIF) for precise cancer survival prediction.
• To address limitations in current multimodal fusion strategies for exploring complex interactions and the capability-efficiency dilemma.

Main Methods:

• Developed a novel multi-shot fusion framework that decomposes fusion into successive stages for progressive integration of pathological and genomic data.
• Introduced affinity-based interactive modules to efficiently generate low-dimensional, discriminative multimodal representations by leveraging comprehensive affinity information.

Main Results:

• The proposed MIF method achieved state-of-the-art performance in cancer survival prediction across various datasets.
• MIF demonstrated superior computational efficiency compared to existing multimodal survival prediction methods.

Conclusions:

• The multi-shot interactive fusion approach effectively integrates multimodal data for precise cancer survival prediction.
• MIF offers a promising solution to the capability-efficiency dilemma in multimodal data analysis for oncology.