抄録
With the development of the clinical database infrastructure, medical knowledge discovery in clinical databases has been actively studied recently. To obtain useful knowledge for medical treatment, it is necessary to apply time-series analysis methods to clinical time-series data. However, such data is nonstationary because of changes in symptom, and the clinical examination intervals of the data are extremely irregular. That makes the application of time-series analysis methods difficult because conventional time-series analysis methods assume stationarity and regularity. Therefore, we herein propose an interpolation method to make the time intervals of clinical timeseries data equal based on recursive regression diagnosis and segmentation. We conducted an evaluation experiment using real clinical examination records of hepatitis and compared the performance of the proposed method to that of a conventional interpolation method based on averaging points in each constant period. The results indicated the potential of the proposed method to segment periods of different symptoms and interpolate the points during each period.We mainly confirmed the segmentation performance in this experiment, and the future work will be the estimation of the whole interpolation performance.
