Artificial intelligence in multimodal data analysis for cancer survival prediction

Area of Science:

• Computational biology
• Medical informatics
• Artificial Intelligence in Oncology

Background:

• Artificial intelligence (AI) is crucial for precise cancer patient survival predictions using multimodal data.
• Integrating diverse data modalities (e.g., imaging, clinical, genomic) presents challenges in alignment and fusion.
• Current AI models struggle with heterogeneous data, impacting accuracy and interpretability in cancer prognosis.

Purpose of the Study:

• To survey existing data alignment and fusion methods for AI-driven cancer survival prediction.
• To analyze recent deep learning studies (2015-2025) on multimodal data integration in oncology.
• To identify challenges and future directions for AI-based cancer survival prediction frameworks.

Main Methods:

• Systematic literature review of 31 multimodal deep learning studies focused on cancer survival prediction.
• Categorization of studies based on alignment and fusion strategies (18 with alignment, 13 without).
• Analysis of data types, alignment techniques, and fusion strategies employed in the reviewed studies.

Main Results:

• Data alignment prior to fusion significantly improves AI model performance and interpretability in cancer prognosis.
• 18 studies demonstrated benefits of alignment followed by fusion, while 13 focused solely on fusion.
• Deep learning is the predominant technique for multimodal data integration in recent cancer survival research.

Conclusions:

• Synchronizing multimodal data before fusion is a key strategy for enhancing AI-based cancer survival prediction.
• Methodological challenges in data alignment and fusion need addressing for clinical relevance.
• Future research should focus on developing robust and interpretable AI frameworks for personalized cancer prognosis.