Affiliations: [a]
Department of Computer Science and Engineering, SCAD College of Engineering and Technology Tirunelveli, Tamilnadu, India
| [b]
Department of Computer Science and Engineering, College of Engineering, Guindy, Anna University, Chennai
Correspondence:
[*]
Corresponding author. A. Essaki Muthu, Assistant Professor, Department of Computer Science and Engineering, SCAD College of Engineering and Technology Tirunelveli, Tamilnadu, India. Email: muthuessaki181@gmail.com.
Abstract: Cataract, a common eye disease, causes lens opacification, which can lead to blindness. Early cataract detection in a privacy-preserving approach has led us to investigate the concept of Federated Learning (FL) and its prominent technique, known as Federated Averaging (FedAVG). Federated learning has the potential to solve the privacy issues by allowing data servers to train their models natively and distribute them without invading patient confidentiality. This research introduces an interactive federated learning framework that permits multiple medical institutions to screen cataract from split lamp images utilising convolutional neural network (CNN) without sharing patient data, as well as grade normal, mild, moderate, and severe cataracts. The CNN is developed based on Modified-ResNet-50 and FedAVG technique could achieve relatively high accuracy. The experimental results demonstrate that the proposed modification reduces the processing time to a greater extent.
