2025
Emotion Recognition × Education Support AI "EMOTIP"
Exhibited at the Osaka-Kansai Expo
: Initiatives to Improve Care Quality and Reduce Turnover Rates in Care Facilities using AI Video Analysis

Background

The care industry faces a chronic labor shortage. In a survey of over 500 caregivers, 15.4% cited "no future prospects" and 14.9% cited "current salary" as reasons for leaving the profession within 3 years, confirming that future uncertainty and working conditions are primary drivers of turnover.
Furthermore, difficulties communicating with users who have trouble expressing emotions, such as dementia patients, were a major source of stress for both users and caregivers.

Challenge

A system is needed to correctly understand the emotions of care recipients and provide appropriate support, thereby reducing resident stress and caregiver burden. Instead of traditional observation and record-based evaluations, there was a need to visualize care quality and support the self-evaluation of care skills, reducing the burden of staff training and lowering turnover rates.

Tasks

Exhibited at the "Future Life Experience" (July 8 - July 14) of the Expo 2025 Osaka-Kansai, promoting a project to entrust everyone's future visions to the "Tree of the Life-Radiating Future" and spread them to the world.

2024
Clinical Trial Success Prediction for Immune Checkpoint Inhibitors
: AI Utilization for Efficient and Effective Drug Development

Background

Drug development takes over ten years and costs hundreds of billions to trillions of yen, with success probabilities dropping from 1 in 15,000 a decade ago to 1 in 30,000 today. In this context, optimizing and accelerating clinical trials and resource allocation has long been a challenge for the industry. Specifically, the success rate for transitioning from Phase II to Phase III is only 24.6%, leaving significant room for improvement.

Challenges

Worked on predicting the success probability of Phase III clinical trials for immune checkpoint inhibitors. By applying the "capturing representations from observed information" concept of world models, we visualized and verified precision improvement based on patient blood and vital data.

2023
Realizing Rentable but Non-Transferable Personal NFTs
: Building a Transparent Data Management System for Medical Data Use

Background

In the modern medical field, data utilization is still underdeveloped. I focused on converting data managed by medical institutions into "data whose value the individual can recognize and utilize." However, this concept requires building mechanisms that ensure both security and transparency. Until now, there were no successful cases of such implementation, leaving numerous technical and operational challenges unsolved.

Challenges

While individuals providing health data want to restrict specific elements and the duration of data sharing, insurance companies require a system that ensures data accuracy while enabling business use.