Revolutionizing Healthcare: Federated Learning for Large-Scale Image Processing in IoT-Integrated Systems

The internet of things (IOT) devices integrated into federated learning (FL) is a revolutionary concept in the fast-paced world of healthcare, especially in medical image processing. In this paper, we propose a novel framework with the enablement of the Internet of Things (IoT)-based medical devices...

Full description

Saved in:

Bibliographic Details
Published in:	2024 5th International Conference on Smart Electronics and Communication (ICOSEC) pp. 432 - 439
Main Authors:	Ravi, Kiran Chand, Vadlamani, Subramanyam M, KumarBasu, Asim, Kumar, Vivek, Mallik, Satadal, Sreelatha, K.
Format:	Conference Proceeding
Language:	English
Published:	IEEE 18-09-2024
Subjects:	Accuracy Cascaded Convolutional Neural Networks Computational modeling Convolution Convolutional neural networks Data Security Diagnostic Accuracy Federated learning Healthcare Image Processing Internet of Things IoT Medical diagnostic imaging Medical Imaging Medical services Privacy Spatial-Variant Convolution Kernel Synchronization Training
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	The internet of things (IOT) devices integrated into federated learning (FL) is a revolutionary concept in the fast-paced world of healthcare, especially in medical image processing. In this paper, we propose a novel framework with the enablement of the Internet of Things (IoT)-based medical devices and Federated Learning to autonomously handle large-scale medical images in an efficient and secure way. The approach adopts Cascaded Convolutional Neural Networks with Spatial-Variant Convolution Kernel (CCSVCK) for improved feature extraction and classification performance. The framework de-centralizes the training process and thus allows cohering to patient privacy and other strict data protection regulations. We used a large dataset of mixed real and synthetic medical images after various preprocessing steps, such as normalization, augmentation, and anonymization, to implement and experiment on all stages of our approach. Training iterations were launched on each local node and contributed model updates to a central coordinator for aggregation. At this point in time, the model was synchronized across all nodes again and iteratively a global model was continuously computed. Evaluation metrics confirmed the robustness of the framework, yielding a commendable accuracy of 97.2%. This novel approach is a blend of not only providing better accuracy in diagnosis but also taking care of the privacy of the information, which has great potential to change health care diagnostics into IoT and Federated Learning.
DOI:	10.1109/ICOSEC61587.2024.10722077