AN ENHANCEMENT TO FUSED AND CASCADED SQUEEZE AND EXCITATION NETWORK FOR PNEUMONIA DETECTION

Abstract

Pneumonia can be caused at any stage of life. In 18% of the cases, infectious diseases lead to pneumonia. Sometimes, such situations may even also lead to death. In order to detect Pneumonia, medical professionals examine lung X-rays and it would be beneficial to develop an automated method for detecting pneumonia so that it can be treated without delay. Convolutional Neural Classification became popular after deep learning algorithms were successfully used to analyze medical images. The use of CNNs for disease analysis has gained much attention. A pre-trained CNN model trained on large datasets yields a great deal of useful information for image classification. The proposed work uses lung X-ray images to visualize two well-known convolutional neural networks, Inception V3 and Squeeze Excitation. Initial phases include preprocessing, feature extraction, and selection of features. The proposed model is found to have 95.39 % accuracy in the tests.