Deep Learning Model for COPD Classification Using XiResNet-50 with Explainable AI

The ongoing development of Deep Learning techniques revolutionizes the Chronic Obstructive Pulmonary Disease (COPD) diagnostic process as well as the analysis of heart conditions such as COVID-19, lung opacity, and other similar lung-related ailments. In this research, we propose a classification model based on ResNet-50 architecture which was enhanced with Explainable AI (XAI) techniques where Shapley Additive Explanations (SHAP) techniques were used to boost the model interpretability. A deep convolutional neural network, ResNet-50 is selected for use because it employs residual networks that improve feature extraction while eliminating the vanishing gradient problem. We utilized transfer learning with ImageNet pre-trained weights to achieve better results while also achieving efficiency. To reduce overfitting, we used augmentation in the form of rotation, scaling, and flipping. Greater prediction accuracy is ensured by the use of Global Average Pooling (GAP) as well as additional fully connected layers in the classification head. The model's training was done using the Adam optimizer with a categorical cross-entropy loss function. Additional model interpretability is provided by SHAP-based heatmaps which show areas that greatly influenced the model's predictions, giving radiologists insight into the offered explanations. The model's quantitative performance is measured with accuracy, the number of iterations needed for convergence, and AUC, the results of which confirm the effectiveness of the model. Combining techniques to explain deep learning models using AI streamlines the gap left between prediction systems and clinical diagnosis systems. This work proposes a new framework for lung disease classification that justifies the trust in AI models thereby enhancing the quality of medical diagnostic systems and its reliability.