9th International Conference on ICT for Sustainable Development

Authors - Khushi Shah, Akshar Patel, Hemant N Yadav
Abstract - The prognosis of a patient depends critically on an accurate diagnosis, but traditional methods are frequently unreliable and yield insufficient data. To improve pneumonia detection, we investigate in this work the automated use of transfer learning models. We carefully trained and validated five well-known convolutional neural network (CNN) models: ResNet152V2, VGG18, ResNet50, InceptionV3, and MobileNetV2. We did this by using a carefully calibrated Chest X-ray dataset. We obtained an impressive accuracy of 97% using an extensive evaluation that included confusion matrices, accuracy and loss graphs, and intermodel comparisons. This breakthrough outperforms conventional methods, highlighting the effectiveness of transfer learning for accurate pneumonia diagnosis. The enhanced diagnostic procedure, which has a specificity of 0.98, promises better patient outcomes by enabling more precise diagnoses, cutting down on pointless sampling, and boosting medical confidence. This study highlights the critical role that transfer learning plays in improving diagnostic accuracy as well as the transformative potential of deep learning methodologies in medical imaging analysis.