HGSP-CNN: Hypergraph Spiking Neural Framework for Accurate Pulmonary Disease Classification using Chest X-Ray Images

Abstract

Pulmonary disease encompasses a diverse range of conditions that affect the respiratory system, including the nerves and organs involved in breathing. These disorders disrupt normal gas exchange, designing inhalation and exhalation difficult. They range from mild and self-limiting illnesses, such as the common cold and catarrh, to severe and life-threatening conditions, like Viral Pneumonia (VP), Tuberculosis (TB), Bacterial Pneumonia (BP) and acute respiratory syndromes like coronavirus disease 2019 (COVID-19). However, it is challenging to detect early-stage pulmonary disease using Computed Tomography (CT) scans and Chest X-Ray (CXR) images because of the similarity of pulmonary abnormalities to surrounding structures, which results in inefficient detections. To overcome this problem, this research proposes the Hypergraph Spiking P system based Convolutional Neural Network (HGSP-CNN) approach for pulmonary disease classification. The proposed HGSP-CNN dynamically constructs feature-dependent hyperedges from intermediate CNN activations while enabling high-order spatial reasoning. The experimental discoveries illustrate that proposed HGSP-CNN approach reaches the optimal accuracy of 99.58% and 99.25% on CXR14 dataset and COVID19 Radiography datasets as compared to the existing methods like ResNet50-V2 and PulmoNet.