Data Sheet 2_AI-Accelerated blockchain architecture for efficient and secure electronic healthcare record exchange.pdf

This work presents the AI-Accelerated Blockchain Architecture (AIBA-EHR), a framework engineered to accelerate transaction processing and enable adaptive security measures suitable for real-world clinical environments.</p>Methods<p>Our approach couples a lightweight Python-based blockchain implementation with asynchronous parallel block validation and machine-learning-driven consensus parameter optimization.

A Proximal Policy Optimization (PPO) reinforcement learning agent anticipates network congestion and dynamically calibrates block size alongside validation difficulty. A risk estimation module built on Temporal Convolutional Networks with Attention Mechanisms (TCN + AM) prioritizes trustworthy nodes and identifies anomalous access behavior. The TCN + AM module employs dilated causal convolutions with dilation rates d ∈ {1,2,4,8}, yielding a receptive field of 31 time steps that fully covers the 20-step access sequences used in this study.</p>Results<p>Experimental evaluation using synthetic EHR datasets comprising 120,000 access events across 15 features reveals a mean confirmation latency of 25.7 s and an average throughput of 209.7 transactions per second (TPS), representing a 40% reduction in confirmation latency and a 25% improvement in throughput relative to fixed-parameter blockchain baselines.