Patent ductus arteriosus (PDA) is a disease commonly found in preterm infants with a short gestation and low weight. Although it occurs in normal infants, it is especially fatal to preterm infants. Therefore, in order to reduce the mortality rate and the probability of future sequelae of infants, early treatment for individual patient is important. Until now, the treatment of PDA was carried out after diagnosis through ultrasound, but if it can be predicted in advance, it will be possible to prepare for treatment. Therefore, we analyze the risk factors for PDA using electronic health records from hospital and examine the PDA probability. These results provide clinical support for medical professionals to classify at-risk patients and prepare for treatment.
Further, we design a diagnostic support system so that medical professionals can directly use and receive medical help. From the point of view of medical professionals (or system users), information of individual patient is delivered to them in several ways so that they can trust machine learning results and apply them to medical processes. Data visualization or interpretation from various perspectives is applied. Such a diagnostic support system will be configured to enable application and support not only for PDA, but also various other diseases.