Abstract

Purpose: This study proposes an engineering design model for a blockchain-enabled collaborative medical information platform that integrates patented AI imaging-diagnosis technology. Research design, data and methodology: Building upon Patent KR10-2604558, the platform incorporates a multi-layered architecture consisting of an AI diagnostic engine, a self-corrective learning loop (ADE), a PBFT-based blockchain integrity module, and an FMEA-driven risk-management framework. A synthetic dataset of 2,000 musculoskeletal ultrasound images was generated to evaluate the structural feasibility of the proposed model. The AI module, developed using a ResNet50 backbone and a four-class Softmax classifier, demonstrated stable self-correction through ADE, which autonomously identified false positives and false negatives and used them to construct a hard-case dataset for selective retraining. Results: The blockchain module—designed with ECC-256 encryption, SHA-256 hashing, and a seven-node PBFT network—successfully ensured immutability, tamper detection, and privacy preservation using Zero-Knowledge Encryption Exchange (ZKEE). FMEA analysis confirmed that risks related to AI misclassification, data integrity, consensus failure, and user input errors could be decomposed into modular risk structures, resulting in a 44.8% reduction in overall RPN.Conclusions: The findings demonstrate that the proposed design model can serve as a technically reliable architecture for AI-driven, legally robust, and privacy-preserving collaborative healthcare data platforms