In a joint research with Dr. Yamamoto and colleagues at Kagawa Prefectural Central Hospital, we designed and implemented an AI model to determine osteoporosis based on cervical X-ray images and patient data.
The "Multi-Modal Deep Learning" which combines X-ray images and text data in medical records, has achieved performance that exceeds the accuracy of conventional deep learning.
This research paper has been published in the Swiss-based academic journal MDPI Biomolecules as "Deep Learning for Osteoporosis Classification Using Hip Radiographs and Patient Clinical Covariates.
We have participated as a technology partner in the development of BrainSuite, a revolutionary brain checkup program that visualizes the health level of the brain and the risk of dementia using machine learning algorithms.
Our work included the integration of the algorithm devised at the Institute of Development, Aging and Cancer, Tohoku University into the application, the connecting of the MRI image handling part and the cognitive function test, and the implementation of the basic part of the online medical interview.
We also facilitate meetings in both English and Japanese for international teams that include foreign engineers and researchers. Throughout the long-term project, our team was involved in the overall design of the application, including cyber security, and we were able to respond to reviews from medical institutions as needed during the development process.
BrainSuite is sold by CogSmart in partnership with Philips Japan, Inc. to various medical institutions. We are confident that this unprecedented brain health checkup using AI will help more people improve their brain health and prevent dementia in the future.
We were engaged in a project to analyze the text information contained in the third-party evaluation report of surgical operations held by the National Cancer Center to determine the key elements of surgical operations and the surgical skills required.
The project processes the large amount of natural language information contained in previous reports and analyses it with a machine learning pipeline we have built. We obtained trends in the keywords included, and automatically categorized each sentence to see which of the specialized evaluation items it fell under.
This project is part of the National Cancer Center's project to evaluate surgical skills, and will be used to improve the quality of surgical care.