Hybrid CNN Model for Pulmonary Disease Detection

This research proposes a hybrid convolutional neural network (CNN) model for detecting various pulmonary diseases using a substantial dataset of Lung CT-Scan images. The hybrid architecture integrates ResNet, DenseNet-121, and InceptionV3 to harness diverse feature extraction capabilities, targeting the identification of diseases like AdenoCarcinoma, Large Cell Carcinoma, Squamous Cell Carcinoma, COVID-19, and Normal cases. The goal is to enhance accuracy and sensitivity in pulmonary disease identification for early diagnosis and intervention.The hybrid CNN model undergoes training on the extensive dataset, utilizing transfer learning techniques to leverage pre-trained weights from ResNet, DenseNet-121, and InceptionV3. The training process fine-tunes the model, enabling it to capture intricate patterns indicative of various pulmonary diseases present in Lung CT-Scan images, with a specific focus on distinguishing between different disease categories.Evaluation on an independent test dataset demonstrates the model's efficacy, exhibiting improved performance compared to individual CNN models. The hybrid CNN model achieves an average accuracy of 98.61% with a loss of 0.0971 in training, and an average accuracy of 81.70% with a loss of 0.4649 in validation. In the test phase, the model attains an accuracy of 84.40% with a loss of 0.4387. Preliminary results suggest that the hybrid CNN model provides enhanced accuracy and sensitivity in pulmonary disease detection, particularly in the classification of AdenoCarcinoma, Large Cell Carcinoma, Squamous Cell Carcinoma, COVID-19, and Normal cases. The amalgamation of various CNN architectures enhances the model's ability to recognize both subtle and prominent patterns associated with different pulmonary conditions. This research contributes significantly to advancing automated diagnostic tools for pulmonary diseases, aiming to facilitate early detection and improve overall healthcare outcomes.