DATE2026.09.14 #Press Releases

A Dynamic Multimodal Survival Prediction Model for Multiple Myeloma

Disclaimer: machine translated by DeepL which may contain errors.

—Integrating Gene Expression, Longitudinal Test Results, and Treatment History—

Summary

A research group led by graduate student Shangru Jia from the Graduate School of Frontier Sciences at The University of Tokyo; Alok Sharma, a full-time researcher at the RIKEN Center for Life Science and Medical Research; Artem Lysenko, an Associate Professor at the Graduate School of Science, University of Tokyo; Professor Tatsuhiko Tsunoda (who also serves as a professor at the University of Tokyo’s Graduate School of Frontier Sciences), among others, have proposed a dynamic multimodal predictive model for multiple myeloma that forecasts residual overall survival based on an observation period of 1 to 18 months following diagnosis. Conventional prognostic stratification for multiple myeloma has been based on staging determined at the time of diagnosis and has not taken into account information obtained over time during treatment. In this study, we integrated gene expression data at diagnosis converted into images using the research group’s proprietary DeepInsight method, clinical and blood test data at diagnosis, and the temporal trends of 10 types of test items and treatment history using deep learning, we were able to accurately predict survival time from the point of predictive analysis onward. It is expected that this method will contribute significantly in the future to improving the accuracy of prognosis prediction for diseases requiring long-term treatment, such as multiple myeloma, by actively incorporating information obtained throughout the course of treatment.

AI-assisted treatment for multiple myeloma

Links

“Genomics with Artificial Intelligence” (August 6, 2019) – DeepInsight Method

Journals