Enhanced Inception Residual CNN for Lung Nodule Detection Using LIDC Dataset

Abstract

Over Several Decades the early identification of lung nodules in computed tomography images is avitalstage in the diagnosis and assessment of lung cancer. The variation in nodule size, shape, texture, and intensity makes automated detection a challenging image-analysis task. In our work an Enhanced Inception Residual Convolutional Neural Network for detecting pulmonary nodules from the Lung Image Database Consortium and Image Database Resource Initiative dataset. The proposed network combines the multi-scale feature learning capability of Inception modules with residual connections, allowing the model to capture both local and broader image characteristics while maintaining effective information flow through the deeper layers. The overall framework includes image pre-processing, augmentation, feature learning, and classification to separate nodule regions from non-nodule regions. Augmentation is incorporated to provide greater variation in the training samples and improve the ability of the network to perform on previously unseen images respectively. The effectiveness of the proposed approach is assessed using accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve. In order to examine the contribution of the enhanced architecture, The performance is compared with conventional convolutional neural network, residual convolutional neural network, and Inception convolutional neural network models. The study aims to develop a dependable computer-aided detection method that can learn relevant characteristics of pulmonary nodules and provide consistent support for computed tomography image interpretation respectively.