A Survey of Methods and Architectures for IoT-Based Breast Cancer Detection with Bayesian Quantized Neural Networks Using Energy-Efficient WSN

Abstract

Breast cancer remains a major global health concern, where early detection significantly improves survival rates. Recent advancements in Internet of Things (IoT)-based healthcare systems have enabled real-time monitoring, remote diagnosis, and efficient data transmission through Wireless Sensor Networks (WSNs). However, challenges such as energy efficiency, computational complexity, uncertainty in diagnosis, and secure data transmission persist. Deep learning techniques, particularly Convolutional Neural Networks (CNNs), have demonstrated high accuracy in detecting breast cancer from medical imaging modalities such as mammography and MRI, achieving accuracy levels exceeding 90%. To address uncertainty in predictions, Bayesian Neural Networks (BNNs) have been introduced, providing probabilistic outputs and confidence measures that improve clinical decision-making. Additionally, quantized neural networks reduce model complexity and energy consumption, enabling deployment on resource-constrained IoT and WSN devices. IoT-based frameworks further enhance system scalability and accessibility by enabling continuous monitoring and data sharing across distributed environments. This survey presents a comprehensive analysis of methods and architectures for IoT-based breast cancer detection systems, focusing on Bayesian quantized neural networks and energy-efficient WSN architectures. It reviews recent developments, compares methodologies, identifies research gaps, and discusses future directions. The study emphasizes hybrid architectures that combine deep learning, uncertainty modelling, and energy-efficient networking to achieve reliable and scalable healthcare solutions.