astPSL: Similarity Learning System Based on Structured and Unstructured Records

Jie Zhang, Wenxin Zhang, Mengfei Kang, Xin Zhang, Lei Zhu, Xinhong Hei

Abstract: Patient similarity learning (PSL) techniques analyze electronic health records (EHRs) to assist with clinical diagnosis and decision-making. However, existing methods often face challenges like incompatibility with EHR systems and inability to handle data updates dynamically. To overcome these limitations, we have developed astPSL, a user-friendly and optimizable PSL system. The astPSL allows users to interact with the system in a customized mode, tailoring the experience to their specific needs. It can continuously optimize the learning results of the PSL technique based on an iterative reinforcement policy, providing users with enhanced learning capabilities. This adaptive optimization approach ensures that the system remains up-to-date and effective as new data becomes available. Our system empowers ordinary users to leverage PSL techniques easily and effectively, which has notable application value and practical significance. By addressing compatibility issues and enabling dynamic updates, astPSL overcomes common limitations of traditional PSL methods, making it a valuable tool for healthcare professionals seeking to leverage patient data for informed decision-making and improved clinical outcomes.