Harnessing MobileNet for Accurate Classification of Respiratory Conditions: An Innovative Methodology

This research explores the application of MobileNet, an efficient convolutional neural network, for the classification of lung illnesses, notably targeting three unique classes: normal lung, lung opacity, and viral pneumonia. The study underscores the importance of adopting MobileNet due to its lightweight architecture, making it particularly appropriate for deployment in mobile and edge computing situations. Through a rigorous methodology encompassing thorough data preparation, the model was trained and tested over 20 epochs, resulting in impressive performance metrics. The model got a training accuracy of 100%, with a validation accuracy of 96.58%, exhibiting its great learning skills and generalization to unknown data. Performance testing further demonstrated good accuracy, recall, and F1-scores across all classes, with values of 92.68%, 94.43%, and 95.43% for the relevant classes, underscoring the model's dependability in correctly identifying lung disorders. The decreasing validation loss, from 1.0557 to 0.3315, suggests strong learning and effective performance on the validation dataset. This study emphasizes the potential of MobileNet to increase diagnostic accuracy in medical imaging, delivering valuable insights into the integration of deep learning technologies in healthcare settings. The findings indicate for ongoing advances in model design and training processes, paving the road for breakthroughs in early detection and treatment of lung illnesses. Ultimately, this research underscores the usefulness of artificial intelligence in updating diagnostic techniques and improving patient outcomes.