Recognition and Classification of Medical Images using Convolutional Neural Network

Abstract

Pneumonia diagnosis is indeed challenging and error-prone to diagnose in chest X-ray pictures as other lung infection transparency is also prominent in the scan. Manually analysing X-rays by different professionals can yield distinct conclusions and inaccuracies. Implementing convolutional neural network (CNN) based models that can correctly categorize specific pneumonia categories can help with targeted treatment. The objective of this study is to utilize a pre-trained CNN variation along with Mobile Net in order to train a data set from an open source comprised of clinical images of an X-ray, followed by a training and validation analysis of the method. By analysing the data, the accuracy and adaptability of the CNN model has been evaluated. The highlights of the study defines that the imbalanced data has been balanced and enhanced the true positive prediction with improved accuracy and precision.