Deep multimodal fusion of patho-radiomic and clinical data for enhanced survival prediction for colorectal cancer patients

1

PRISM-CRC is a deep learning framework that integrates histopathology, radiology, endoscopy, and clinical data to enhance colorectal cancer diagnosis and prognosis.
2

The model achieved a concordance index of 0.82 for predicting 5-year disease-free survival and an AUC of 0.91 for identifying microsatellite instability status.
3

PRISM-CRC outperformed single data type models, providing more granular risk stratification than the traditional TNM staging system.
4

The framework has clinical implications for personalizing treatment, particularly for high-risk Stage II colorectal cancer patients.
5

Limitations include performance decrease due to domain shift and classification errors, necessitating future prospective trials for validation.

Npj Digital Medicine

1