ScholarWorks@성균관대학교

초록

With the growing demand for accurate yet effortless health monitoring at home, most current approaches to stool analysis rely on self-reported diaries that are prone to recall bias and low adherence. Here we present a fully passive alternative: the Precision Health Integrated Diagnostic (PHIND) system, a smart-toilet-based platform that enables automated defecation monitoring without requiring users to alter their daily routines. By integrating optical and pressure sensors with cloud-based convolutional neural networks, the PHIND system classifies stool form according to the Bristol Stool Form Scale and records key defecatory parameters, including total event time, defecation duration and time to first stool drop. The protocol proceeds in three principal stages: (1) assembling and mounting the hardware onto a conventional toilet; (2) training convolutional neural network models for stool classification and event detection; and (3) image acquisition and deploying cloud infrastructure for real-time analysis, data storage and visualization. Compared with traditional methods that depend on user-reported stool diaries, PHIND provides objective, near real-time data free from recall error, enabling more reliable early detection and long-term management of gastrointestinal conditions. Researchers and clinicians can expect high classification accuracy and robust, longitudinal insights into defecation patterns. The complete protocol-from hardware setup to system validation-can typically be completed within 2 d, excluding printed circuit board manufacturing, which generally requires up to 15 d depending on the manufacturing provider.

키워드