Development status

Development of Medical Solutions

The application of AI in the medical field is a medical theme with great potential, but there are still few examples of its practical application. Based on 1) a perspective that emphasizes understanding of medical needs and development in the medical field, 2) a network with many physicians and departments, and 3) experience and know-how accumulated through investigator-initiated clinical trials, we are working as a hub that links physicians at medical institutions, IT vendors with AI technology, and pharmaceutical/health tech companies for industrialization. We are working to build an ecosystem that connects AI research to business in the medical field. Since we can conduct clinical trials in accordance with the Pharmaceutical Affairs Law, it will be possible for us to develop full-scale medical solutions (diagnosis and treatment) that can be used in actual clinical practice.

Respiratory Function Diagnosis

Although spirometry is the most important test for respiratory diseases and respiratory function, it is not widely used. This is because it is difficult for non-specialists to interpret the results (flow volume curve). In collaboration with Kyoto University and NEC Solution Innovator Ltd, we are developing an AI algorithm to diagnose respiratory functions and diseases from spirometry test results (flow volume curves). In July 2020, we signed a joint development and commercialization agreement (license agreement) with CHEST M.I. Inc., a leading company in spirometry. Since development of an initial AI model capable of differential diagnosis of various respiratory diseases has been successfully completed, we received a milestone in October 2021 based on the agreement with Chest M.I. Inc. We plan to improve the prediction accuracy by improving the number and quality of data in order to develop it for commercialization.

Support for chronic dialysis system

Hemodialysis is the major treatment for kidney failure. Normally, a dialysis hospital manages dozens of patients with one doctor and several nurses and clinical engineers, but the human resources are not sufficient, and the occurrence of complications such as hypotension during dialysis consumes scarce human resources and also adversely affects the life prognosis of patients. We are developing AI algorithms to predict hypotension events in advance with Tohoku University, University of Tokyo, St. Luke’s International University, several private medical institutions, and NEC Corporation (NEC). We are able to collect a large amount of high-quality medical data of dialysis from our medical network. In the future, we will further increase the amount of medical data (hundreds of thousands of dialysis treatments) and improve the prediction accuracy by upgrading the analysis engine to enable reinforcement learning for individual patients. We have concluded a collaboration agreement with Nipro Corporation, a global hemodialysis medical equipment supplier, in May 2021 with a view to industrialization. Furthermore, we plan to improve the prediction accuracy by improving the computer algorithm for predicting hypotension during dialysis of individual patients. The performance of the AI system is being tested in a clinical setting in 2022, once we complete the analysis with a large amount of data.

Support for diabetes treatment

Insulin injection therapy is necessary for strict control of blood sugar levels in diabetes mellitus. Since the safety margin of insulin therapy is narrow and overdose can cause the side effect of hypoglycemia, the optimal type and dosage must be set for each patient. This requires the knowledge and experience of a specialist for diabetes, and it is difficult for a non-specialist to set the insulin dose. In collaboration with Tohoku University and NEC Corporation, we have developed an AI algorithm that can predict insulin dosage within an error of a few units. In the future, we will further increase the amount of data and improve the AI algorithm in order to increase the accuracy. Currently, we are focusing on AI algorithms that can predict insulin dosage, but in the future, we plan to develop AI algorithms that can select the optimal treatment from among insulin and many oral diabetes drugs, taking into account the medical condition and living environment of each diabetic patient.

Assessment of speaking function and swalloing function

In today’s aging society, dysphagia is considered to be a major cause of pneumonia, but there is no simple method for assessment of swallowing function. The organs used for ‘speaking’ and ‘swallowing’ are largely common, and we have focused on the possibility of predicting the swallowing function from the evaluation of the speaking function. We are working on the development of an AI algorithm to evaluate the swallowing function by analyzing the difference in frequency between the pronunciation of healthy people and that of patients.

Pediatric Developmental Disorders (Dyslexia) Diagnosis

Developmental disabilities are important disorders that impair the social life of children, including autism spectrum disorder (ASD), attention deficit/hyperactivity disorder (ADHD), and learning disabilities. With early treatment and proper education, those with developmental disorders can lead to a social life similar to normal people. The prevalence of dyslexia, one of the learning disabilities, is as high as 2% in children, and as a result of the time consuming and burdensome process of reading aloud, it can lead to poor school performance and truancy. In collaboration with Tohoku University and NEC Corporation, we are developing a system to evaluate dyslexia using AI, based on data from children reading letters and sentences aloud. We have started to collect data from the developmental disability survey of 8-year-old children at the Tohoku Medical Megabank Organization, which started in October 2021.