A web-based Probabilistic Medical Computation platform based on Bayesian network

Objective The objective of this study was to develop a medical service platform utilizing Bayesian networks for medical modeling and risk inference, with minimal configuration requirements. The platform was designed to provide accurate and efficient disease diagnosis and treatment plans for medical professionals. Methods The platform consists of four layers: database layer, Bayesian network construction layer, Bayesian network inference layer, and client layer. The database layer accepts user data uploads. The Bayesian network construction and inference layers are implemented for Bayes network learning and inference based on the bnlearn package of Python. The client layer allows users to define Bayesian network learning parameters and presents inference results in the form of bar charts. Results The code for our prototype system is available on Github (https://github.com/rose1203/BayesNet-platform-for-medical-computation.git). Our user-friendly and flexible platform allows professionals and IT experts to easily configure appropriate parameters for network structure and parameter learning. User-specified Bayesian networks can be saved for inference. Additionally, the platform supports data visualization for medical staff, which provides doctors with an intuitive understanding of patients' potential disease progression. Based on this information, doctors can formulate corresponding treatment plans and care measures. Conclusion Our interactive online platform, which is based on Bayesian networks, provides healthcare professionals and researchers with a valuable tool to make informed decisions. Combined with models based on real-world data and individual patient cases, our platform can promote personalized healthcare and enhance the quality of healthcare services.