ADAPT-HealthNet: Adaptive Deep Reinforcement Learning with Blockchain-Integrated Chaotic Encryption for Secure and Energy-Efficient Medical Image Transmission in IoT-WSNs

The proliferation of Internet of Things (IoT) and
Wireless Sensor Networks (WSNs) in healthcare has revolutionized remote monitoring and telemedicine. At the same time it introduces significant challenges in secure, energy-efficient transmission of sensitive medical images. This paper proposes ADAPTHealthNet, a novel multi-layer framework that integrates shiftinvariant deep convolutional neural networks (CNNs) for preprocessing, chaotic map-based encryption enhanced by machine learning (ML)-driven key selection, a lightweight blockchain for tamper-proof logging, and deep reinforcement learning (DRL)
for adaptive resource optimization.
Justification for integration (addressing reviewer comment
1): CNN reduces bandwidth, chaotic+ML encryption provides
lightweight security, blockchain ensures integrity, and DRL
dynamically balances the security–energy–fidelity trade-off. Ablation studies (Section V) confirm that removing any module
degrades at least one metric by 20–35%.
By dynamically tuning encryption parameters and routing
strategies based on real-time network states, ADAPT-HealthNet
achieves up to 40% energy savings while maintaining highfidelity image reconstruction (PSNR > 48 dB) and robust security.
Evaluations on a simulated 50-node WSN (with Raspberry
Pi 4 prototype validation) using full-resolution ChestX-ray14,
MRI, and CT datasets demonstrate superior performance over
state-of-the-art baselines, with average energy consumption of
40.96±1.2 mJ per transmission (95% CI) and 100% blockchain
integrity. This framework paves the way for scalable, privacypreserving healthcare IoT deployments.