Challenges of CAD for thorax diseases including Covid-19 by using artificial intelligence

Abstract

Artificial Intelligence(AI) is simulation of human intelligence in machines. It is programmed such that it can think as human and perform actions or take appropriate decision. In current covid-19 pandemic, its important to diagnosis more asymptomatic people to save their life. There various diseases related to thorax such as pneumonia, lung cancer, COPD(Chronic Obstructive Pulmonary Disease). Its leading death cause in world. Even fetus are also effected by pneumonia from birth times. The remote area people also can be saved by proper diagnosis on time by using CAD(Computer Assisted Detection). There is some challenges in training of algorithm in AI to give more accuracy. In this paper those issues such as class imbalance, multi task and data size are discussed with solutions for each problem. Different diseases, which look similar by radiologist can be detected in early stage. The pre-processing and finetuning of thorax x-ray is done before applying to CNN(convolutional neural network). Loss functions are calculated with proper weightage value. So that algorithm work even in small training set. © 2021 Author(s).