Diagnostic Classification of Lung Disease with Chest X-Ray Images Using SNN

Abstract

Some of the lung diseases that affect respiratory and pulmonary functions includes atelectasis, pulmonary infiltrate, pneumonia and pneumothorax. The proposed system is a novel classification method of the lung diseases atelectasis, pulmonary infiltrate, pneumonia, pneumothorax and healthy lungs from X-ray images using SNN. Spiking Neural Network is a type of ANN (Artificial Neural Network) in which information processing in neural nodes and communication between neurons is based on the exchange of spikes. Here, at the initial stage, data augmentation is used for increasing the number of datasets. Then in the preprocessing stage, first the images are filtered using a bilateral filter. Next the enhancement technique called Contrast Limited Adaptive Histogram Equalizer (CLAHE) is used to avoid excessive noise enhancement and minimizes edge shadowing effect. Then the images are reshaped to their respective size. After the preprocessing stage, the resized images are then fed into the Spiking Neural Network (SNN) architecture for extracting the features from the images. Then, the generated features or vectors go through XGBoost and Random forest classifiers for classification purposes. The proposed method has an accuracy of 98%.