English 
Fujitsu Ltd. announced on the 6th that by combining the Hirosaki Health Checkup Causal Network, which integrates and manages big data on approximately 3,000 health checkup results over a 20-year period for residents of the Iwaki district of Hirosaki City, Aomori Prefecture, developed by Kyoto University and Hirosaki University, with Fujitsu’s causal decision-making support technology, which it developed as a core technology for its AI service “Fujitsu Kozuchi,” it has become possible to derive causal relationships in the health and medical field even with limited data. The Hirosaki Health Checkup Causal Network was created by a research group at Kyoto University, including Hirosaki University, who applied their own Bayesian network technology to the ultra-multi-item health big data obtained by Hirosaki University COI-NEXT in the Iwaki Health Promotion Project Health Checkup, to estimate the causal relationships between items as a network, creating a highly reliable causal graph. In addition, the causal knowledge transfer technology, a new function of the causal decision-making support technology developed by Fujitsu, can transfer existing knowledge of causal relationships. By combining these, for example, in the field of health and medicine, even if sufficient reliable data cannot be collected, it is now possible to derive causal relationships by repurposing knowledge from the Hirosaki Health Examination Causal Network.
Fujitsu explains that while data-based decision-making has become widespread in fields such as management, medicine, sports, and manufacturing, there are many cases where sufficient data cannot be collected. For example, in recent years, as interest in health management has increased, data analysis has become important for understanding employees’ health conditions and taking effective measures, but companies with a small number of employees have difficulty securing sufficient data, and they have issues identifying health issues and planning measures. Fujitsu has established research bases at universities in Japan and overseas, and is promoting the Fujitsu Small Research Lab initiative, in which the company’s researchers stay at universities full-time or for long periods of time while engaging in industry-academia collaboration activities. The joint research course between Kyoto University and Fujitsu, the Large-Scale Medical AI Course (Fujitsu Research Lab), has been researching and developing new AI technologies to solve problems in the field of health and medicine. Causal decision-making support technology is a technology that supports decision-making by proposing optimal measures based on causal relationships estimated using multiple data sets.
The newly developed causal knowledge transfer technology first converts the knowledge of causal relationships obtained from existing data into a causal knowledge graph. Then, by identifying the causal structures that can be transferred from the causal knowledge graph at a fine level according to the data distribution, it is possible to transfer the knowledge even when the items and abstraction levels are different. This makes it possible to derive highly reliable causal relationships by transferring the knowledge of causal relationships derived from existing data, even when data is insufficient. By combining this technology with the Hirosaki Health Examination Causal Network, a highly reliable causal graph obtained from existing health examination data of various regions and age groups, it becomes possible to infer causal relationships in the health and medical field even when there is insufficient data. For example, when inferring causal relationships in the “Sleep Health and Lifestyle Dataset,” which is open data on sleep and lifestyle, using the causal knowledge transfer using the Hirosaki Health Examination Causal Network, it was possible to derive more valid causal relationships compared to when the Hirosaki Health Examination Causal Network was not used. For example, if the Hirosaki Health Examination Causal Network is not used, an unnatural causal graph would be derived that shows that age and gender are the causes of insomnia.
こちらもお読みください： ザパス、AIを活用して個人クリニックの集患を効率化するサービス「ClinicHub」を開始
However, by using the Hirosaki Health Examination Causal Network, such unnatural relationships are eliminated, and the reasonable result that sleep time and quality directly affect insomnia is derived. Fujitsu has begun offering a trial environment for health-related corporations that allows them to try out the causal decision-making support technology in combination with the Hirosaki Health Examination Causal Network. Fujitsu explained that in the future, it will continue to use the knowledge gained through the provision of the trial environment to improve the accuracy of the technology and expand its functions, thereby contributing to the further development of AI technology in the health and medical field and to the health management of more companies. In addition, the causal decision-making support technology is a general-purpose technology that can be used in a variety of fields, and it will be applied to financial and non-financial information, for example, to contribute to decision-making support for a wide range of companies.Fujitsu Ltd. announced on the 6th that by combining the Hirosaki Health Checkup Causal Network, which integrates and manages big data on approximately 3,000 health checkup results over a 20-year period for residents of the Iwaki district of Hirosaki City, Aomori Prefecture, developed by Kyoto University and Hirosaki University, with Fujitsu’s causal decision-making support technology, which it developed as a core technology for its AI service “Fujitsu Kozuchi,” it has become possible to derive causal relationships in the health and medical field even with limited data.
The Hirosaki Health Checkup Causal Network was created by a research group at Kyoto University, including Hirosaki University, who applied their own Bayesian network technology to the ultra-multi-item health big data obtained by Hirosaki University COI-NEXT in the Iwaki Health Promotion Project Health Checkup, to estimate the causal relationships between items as a network, creating a highly reliable causal graph. In addition, the causal knowledge transfer technology, a new function of the causal decision-making support technology developed by Fujitsu, can transfer existing knowledge of causal relationships. By combining these, for example, in the field of health and medicine, even if sufficient reliable data cannot be collected, it is now possible to derive causal relationships by repurposing knowledge from the Hirosaki Health Examination Causal Network. Fujitsu explains that while data-based decision-making has become widespread in fields such as management, medicine, sports, and manufacturing, there are many cases where sufficient data cannot be collected. For example, in recent years, as interest in health management has increased, data analysis has become important for understanding employees’ health conditions and taking effective measures, but companies with a small number of employees have difficulty securing sufficient data, and they have issues identifying health issues and planning measures. Fujitsu has established research bases at universities in Japan and overseas, and is promoting the Fujitsu Small Research Lab initiative, in which the company’s researchers stay at universities full-time or for long periods of time while engaging in industry-academia collaboration activities.
The joint research course between Kyoto University and Fujitsu, the Large-Scale Medical AI Course (Fujitsu Research Lab), has been researching and developing new AI technologies to solve problems in the field of health and medicine. Causal decision-making support technology is a technology that supports decision-making by proposing optimal measures based on causal relationships estimated using multiple data sets. The newly developed causal knowledge transfer technology first converts the knowledge of causal relationships obtained from existing data into a causal knowledge graph. Then, by identifying the causal structures that can be transferred from the causal knowledge graph at a fine level according to the data distribution, it is possible to transfer the knowledge even when the items and abstraction levels are different. This makes it possible to derive highly reliable causal relationships by transferring the knowledge of causal relationships derived from existing data, even when data is insufficient. By combining this technology with the Hirosaki Health Examination Causal Network, a highly reliable causal graph obtained from existing health examination data of various regions and age groups, it becomes possible to infer causal relationships in the health and medical field even when there is insufficient data. For example, when inferring causal relationships in the “Sleep Health and Lifestyle Dataset,” which is open data on sleep and lifestyle, using the causal knowledge transfer using the Hirosaki Health Examination Causal Network, it was possible to derive more valid causal relationships compared to when the Hirosaki Health Examination Causal Network was not used. For example, if the Hirosaki Health Examination Causal Network is not used, an unnatural causal graph would be derived that shows that age and gender are the causes of insomnia .
しかし、弘前健診の因果ネットワークを用いることで、そのような不自然な関係が排除され、睡眠時間や睡眠の質が不眠症に直接影響するという合理的な結果が導き出されます。 富士通 富士通は、健康関連企業向けに、弘前健診因果関係ネットワークと組み合わせた因果関係意思決定支援技術を試用できる環境の提供を開始しました。富士通は、今後もトライアル環境の提供を通じて得られた知見を活用し、本技術の精度向上や機能拡充を図ることで、健康・医療分野におけるAI技術のさらなる発展と、より多くの企業の健康経営に貢献していくと説明。また、因果関係意思決定支援技術は、様々な分野で活用できる汎用的な技術であり、例えば財務情報や非財務情報にも適用することで、幅広い企業の意思決定支援に貢献してまいります。
ソース ヤフー
日本語