Machine Learning-Enhanced Architecture Model for Integrated and FHIR-Based Health Data

The widespread fragmentation of patient information across disparate systems and the absence of standardized integration mechanisms hinder efficient and comprehensive medical diagnostics. To overcome these limitations, this work presents an architecture model that supports physicians in the diagnostic process, combining clinical and socio- health information (patients’ medical history) with extracted data from diagnostic reports and images. This architecture allows the identification of risk assessment related to a clinical condition and displays only the necessary information for diagnosis, through the definition of a Decision Support System by leveraging the integration of data from diagnostic images, patient-collected data, and data from heterogeneous sources. Furthermore, the architecture includes the standardization of retrieved and processed information using the international HL7 Fast Healthcare Interoperability Resources (FHIR) standard to enable full integration with Health Information Systems (such as Electronic Health Records and Telemedicine Systems). In this context, a case study concerning the clinical condition of breast cancer is described to demonstrate the functionalities of the architecture, and an AI-based Risk Assessment is performed using ultrasound images. We demonstrate the capabilities of the architecture through a patient-centered mobile Android Application specifically developed for this purpose.