A Visual-Modified Circular Dilated Convolutional Neural Network with Draco Lizard Optimizer Framework for Automated Lung Disease Detection Using CT ImagesOpen Access

Abstract

Lung Disease is one of the prominent causes of death, hence the need for its early diagnosis and accurate detection via CT images. Conventional approaches to diagnosis may have limitations in identifying intricate patterns in lung disorders; hence, the possibility of diagnosis may suffer, leading to delayed medical attention. To mitigate these limitations, the research proposes the Visual-Modified Circular Dilated Convolutional Neural Network with Draco Lizard Optimizer (V-CDCNNet-DLO) for efficient classification of lung disorders. CT images from the LIDC-IDRI dataset are first preprocessed using the Adaptive Self-Guided Loop Filter (AS-GLF) to enhance image quality. The research utilizes the V-CDCNNet for feature extraction and classification, with the help of the weights provided by the Draco Lizard Optimizer (DLO) to improve its performance. The research has demonstrated excellent performance with a high level of accuracy of 99.27%, precision of 99.20%, recall value of 99.15%, and error level of 0.73%. This model establishes a reliable framework for early lung disease detection.