Medical Data Deep Learning Speical Team

The Medical Data Deep Learning Special Team has been active since October 2017 and entered its eighth year in 2025. Initially operating under the Medical Innovation Hub Program, it later joined the Advanced Data Science Program in R-IH before becoming part of the newly established Predictive Medicine Special Project within IMS in April 2025.

The team conducts research on deep learning applications in the medical and healthcare fields. This year, a paper on the application of self-supervised pre-training using synthetic data was peer-reviewed and accepted for publication in Machine Learning for Health, a top conference in medical machine learning. The team is also participating in projects to develop Japanese large language models for medical use, funded by SIP (January-March) and NEDO (April-December), contributing to the creation of domestically developed large language models.

In medical AI research, balancing cross-border data sharing with privacy protection is a critical challenge for all nations, prompting exploration of various international collaborations. This year, we held a two-day international symposium with the French government and research institutes, fostering vigorous discussions.

Additionally, we conducted research on mathematical approaches in basic medical science (Miyasaka et al. Neuron), broad discussions on “AI for Science,” and the application of AI for preventing sports injuries. These works resulted in numerous invited talks and patent applications.