2026-10-13T11:08:03Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33962767 2026-09-22T06:22:41Z category_183 portal_316 item_type_3 month_year_09_2026

Data Sheet 2_AI-Accelerated blockchain architecture for efficient and secure electronic healthcare record exchange.pdf Nagaraj Segar (25088254) Vijayarajan Vijayan (25088257) Distributed Computing anomaly detection artificial intelligence blockchain cybersecurity electronic healthcare records performance optimization proximal policy optimization python implementation Introduction<p>While blockchain integration within Electronic Healthcare Record (EHR) platforms has strengthened data integrity and patient privacy, practical deployment continues to suffer from elevated latency and inflexible processing mechanisms. 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. The TCN + AM anomaly detector achieves an F1-score of 0.9942, precision of 0.9953, recall of 0.9931, and AUC of 0.9950 across five-fold cross-validation (mean ± std: 0.9935 ± 0.0018), outperforming all sequential and tabular baselines on classification metrics.</p>Discussion<p>These findings indicate that the strategic fusion of artificial intelligence with distributed ledger technology can simultaneously achieve operational efficiency and robust security, enabling real-time healthcare data exchange without compromising patient privacy across decentralized infrastructures.</p> 2026-09-22T06:22:41Z Dataset Dataset 10.3389/fbloc.2026.1905158.s001 https://figshare.com/articles/dataset/Data_Sheet_2_AI-Accelerated_blockchain_architecture_for_efficient_and_secure_electronic_healthcare_record_exchange_pdf/33962767 CC BY 4.0