Daijiworld Media Network – Bengaluru

Bengaluru, Oct 11: In a major breakthrough for precision oncology, researchers have developed a multimodal artificial intelligence model that significantly improves cancer prognosis prediction by integrating pathology images, genomics, and clinical data.

Named MICE (Multimodal data Integration via Collaborative Experts), the AI model demonstrated superior performance and efficiency across 30 cancer types, marking a step forward in data-driven cancer care.

Researchers trained and validated the system using data from 11,799 patients, overcoming a key limitation faced by traditional AI models that struggle to process diverse datasets. MICE uses multiple specialized “expert” modules to extract both cancer-specific and shared biological insights, combining contrastive and supervised learning to identify prognostic patterns across different cancers.

The model outperformed existing systems, improving the concordance index (C-index) by 3.8%–11.2% in internal cohorts and 5.8%–8.8% in independent validation sets. Even with limited datasets, MICE maintained robust accuracy, highlighting its potential for use in rare cancer cases where data availability is low.

Experts say the model’s scalability and data efficiency make it ideal for clinical environments, where it could help doctors assess prognosis more precisely and design personalized treatment plans. Researchers believe MICE could soon be integrated into clinical decision-support systems, paving the way for more targeted and effective cancer therapies.