A Customized 3D CNN Integrated with Convolutional Block Attention Module for Precise Diagnosis of COVID-19 and Pneumonia from CT Scan Images

From 2020 to 2022, COVID-19 emerged as a widespread global crisis impacting public health and the global economy. Simultaneously, conventional Pneumonia continues to pose a significant threat to global mortality rates, particularly among children. Notably, the pulmonary ramifications caused by COVID-19 and Pneumonia exhibit striking resemblance, leading to analogous clinical complexities. Therefore, expedient and accurate detection methodologies are imperative to mitigate the mortality toll and overarching consequences of both COVID-19 and Pneumonia. Conventional diagnostic techniques frequently experience delays and errors, which have an impact on patient outcomes. This study developed a tailored 3D Convolutional Neural Network (CNN) with the Convolutional Block Attention Module (CBAM) to classify COVID-19 and Pneumonia precisely from CT images. The 3D CNN structure effectively captured spatial correlations among CT slices, enhancing the model’s ability to detect complex patterns indicative of COVID-19 and Pneumonia. Furthermore, CBAM enhanced feature discrimination by adaptively recalibrating channel and spatial attention, allowing the model to focus on salient features in CT images. The proposed system adeptly learned nuances from CT scans by carefully examining illness-specific properties. Despite several challenges, the proposed investigation got an impressive f-1 score of 98% and an exceptional accuracy rate of 98.29%. This accomplishment highlights the enormous potential of the proposed methodology to transform respiratory problem patient care and diagnostic procedures.