Abstract
Pneumonia is a severe respiratory ailment that may be caused by viruses, fungus, and bacteria. Pneumonia causes the accumulation of water, purulent material, or other fluids in the air sacs (alveoli) of the lungs. A delay in the identification of pneumonia may be life-threatening. Nevertheless, the limited radiation levels used for diagnosis result in unreliable detection, posing a significant obstacle in the field of pneumonia detection in healthcare. Transfer Learning (TL) may revolutionize healthcare by offering an effective method for differentiating between normal and pneumonia patients. Therefore, we have proposed a TL utilizing model for predicting pneumonia. Our experiment uses 5856 highly imbalanced chest X-ray images to fit different vision models such as Xception, VGG16, and ResNet152V2 using TL approach. After training, our model performs exceptionally well on the X-ray dataset, achieving an accuracy of 80.45, 80.77, and 83.17% respectively. However, the results show that the ResNet152V2 performs exceptionally well as compared to other models. Also, it achieves a precision and recall score of 79.87 and 97.69% respectively. The results exemplify the potential of our framework to help pulmonologists and physicians make rapid diagnoses of pneumonia patients by providing accurate and fast pneumonia classification.