An integrated community care system is a medical and care system for the area residents by collaborating various mission of the community. To understand the life environment and to analyze the hospital demand in the future area will help plan an effective care system. It shows how to predict of the hospital demand for the based on public statistical data such as population, health survey, and the number of patients.

For therapeutic decision making related to lower urinary tract symptoms (LUTS) in men, the distinction between detrusor underactivity (DU) and bladder outlet obstruction (BOO) is crucial. However, currently, accurate diagnosis of DU and BOO can only be made by pressure flow study (PFS), which is invasive and complex. To address this problem, this study focuses on the use of uroflowmetry (UFM) waveforms, which can be obtained in a non-invasive manner. More specifically, we construct a couple of one-dimensional convolutional neural networks (CNNs) that can estimate a patient's bladder contractility index (BCI) and bladder outlet obstruction index (BOOI) based on the patient's UFM waveform and classify the patient into one of four categories: normality, DU, BOO, and the combination of DU and BOO. The experimental results show that the constructed CNNs make the diagnosis more flexible and have a diagnostic accuracy comparable to that of a previous approach by human experts.

In medical data sets, there are many types of sequences containing the meaningfulinformation. Sequences of medical orders, specimen inspections, and personal physiological measurementdata are good examples. We applied the sequential pattern mining for them to derive theuseful information for medical workers. However, there still are a number of subjects to be tackled.In this paper, we show some examples of applying the sequential pattern mining for medical data,and discuss the subjects based on the examples.

Although Cisplatin is an effective anti-cancer drug, it is known that there is a risk of Cisplatin-induced acute kidney injury (Cis-AKI). In this study, machine learning-based prediction models were built to detect early-stage Cis-AKI. Our prediction models are designed for the electric health records of individual patients outside the intensive care unit. In the experiments, some of our models achieved moderate predictive accuracy for Cis-AKI, and their predictions visualized by SHapley Additive exPlanations (SHAP) or their decision rules coincide with the medical insights on Cis-AKI.

Convolutional neural networks (CNNs) have been adopted as standard deep learn- ing models in medical image analysis owing to their ability to automatically extract high-level features from training images. Recently, Vision Transformer (ViT) models have been proposed, which implement the Transformer architecture originally developed for natural language process- ing. Given their high predictive performance, we built a couple of ViT models to detect kidney cancer based on computed tomography (CT) images. Experimental results show that our ViT models outperformed conventional CNNs in terms of detection accuracy with various types of CT images. Moreover, we visualized the attention maps of our ViT models to help understand the basis for their detection output.