FedQCNN: A Privacy-Preserving Federated Quantum Convolutional Neural Network for Retinal Image Classification

Abstract

Quantum machine learning (QML) provides the opportunity for the success of an automated medical diagnostic system due to the effective representation of the solution space by quantum entangled states and faster optimisation through quantum superposition. Preserving medical data privacy is crucial while implementing an intelligent and efficient medical service provider system. The federated machine learning model is not only rich in diversified model experiences but also helps to protect patient data privacy. This paper proposes a secure, intelligent Internet of Healthcare Things (IoHT) application with a federated quantum convolutional neural network (FedQCNN) to classify medically significant retinal image patches. Following the gradient descent algorithm, we executed the model optimisation and performed the global aggregation using the weighted average of the local models' parameters. The proposed system achieves an evaluation accuracy of 96.8% on E-Ophtha retinal image dataset. We suggest over-the-air distributed machine learning with wireless multiple access channels to significantly save the scaled-up radio resource requirements for massive connectivity to ultra dense IoT devices. The research emphasises the significant progress of federated quantum machine learning and its prospects in the coming decades.