A Machine Learning Based Approach for Estimation of the Lung Affectation Degree in CXR Images of COVID-19 Patients

The effectiveness of the treatments applied to patients with COVID-19 in serious and critical condition admitted to intensive care units is a necessary element to draw up the strategies and protocols to follow in each particular case. An automatic index that allows to quantify the degree of affectation produced by the disease in the lungs from X-ray images of the thorax has not been investigated so far. The work presents a method for estimation of a lung affectation index in chest X-ray images in patients diagnosed with COVID-19 in an advanced stage of the disease. The index is obtained from a method that combines image quality evaluation, digital image processing and deep learning for lung region segmentation. This method is capable of facing the problem of very diffuse borders due to the notable effects that COVID-19 patients in serious or critical condition have. The subsequent step of our proposal consist in the classification of the previously segmented image into two classes (healthy region, affected region) establishing the relationship between the number of pixels of each class. The results achieved in the experiments on images of healthy and affected by COVID-19 patients showed high values of sensitivity and specificity.