Optimized Deep Feature Selection for Pneumonia Detection: A Novel RegNet and XOR-Based PSO Approach

TL;DR

This paper introduces a novel XOR-based PSO method for deep feature selection in Pneumonia detection using RegNet, achieving high accuracy with minimal computational complexity and providing accessible source code.

Contribution

It presents a simplified XOR PSO algorithm for deep feature selection in CNNs, enhancing accuracy and efficiency in Pneumonia detection.

Findings

Achieved 98% accuracy in Pneumonia detection

Selected 163 features from RegNet model

Proposed method has minimal hyperparameters and computation time

Abstract

Pneumonia remains a significant cause of child mortality, particularly in developing countries where resources and expertise are limited. The automated detection of Pneumonia can greatly assist in addressing this challenge. In this research, an XOR based Particle Swarm Optimization (PSO) is proposed to select deep features from the second last layer of a RegNet model, aiming to improve the accuracy of the CNN model on Pneumonia detection. The proposed XOR PSO algorithm offers simplicity by incorporating just one hyperparameter for initialization, and each iteration requires minimal computation time. Moreover, it achieves a balance between exploration and exploitation, leading to convergence on a suitable solution. By extracting 163 features, an impressive accuracy level of 98% was attained which demonstrates comparable accuracy to previous PSO-based methods. The source code of the proposed method is available in the GitHub repository.