Fig 1.
Patient-Centric Health Data Management Architecture.
The architecture is composed of four functional components: (1) FHIR-compliant data integration for ingesting imaging data and annotations into structured formats; (2) blockchain-based consent management using smart contracts and decentralized identifiers (DIDs); (3) the BioWallet client for patient-controlled access and authorization; and (4) off-chain encrypted storage for safeguarding sensitive data. This pipeline supports secure, standards-based exchange of medical images and metadata.
Table 1.
Comparison of our work with existing literature.
Fig 2.
Workflow for transforming medical imaging files into standardized FHIR resources for AI applications.
Imaging files in DICOM or PNG format are converted through a DICOM-to-FHIR process, followed by manual annotation. The resulting FHIR resources enable standardized data representation for future reuse in AI pipelines. This framework was tested in a simulated ICU setting where clinicians and researchers queried longitudinal datasets based on temporal alignment between imaging, ventilation, and laboratory results. Access control was governed through the BioWallet, allowing simulated patients to grant or revoke access to subsets of their data. These validation activities highlight the feasibility of integrating such a system into clinical research workflows without compromising data sovereignty or privacy.
Table 2.
Cohort overview (ICU ARDS-COVID19).
Fig 3.
Mockups for the application.