Andrews' curve and texture ellipsoid for the detection of preumonia using chest X- ray

Abstract

Patients with lung ailments making their first visit to a medical doctor usually have their chest X-ray taken. At this stage the possibility of misinterpreting the X-ray image is highly plausible. There is a need to develop simple graphical methods to aid in initial screening of patients. Earlier studies showed that a method using Andrews' curve was found to be incapable of differentiating pneumonia (PNEU) from pulmonary tuberculosis (PTB) and lung cancer (LC). As a consequence, this study explored the potential of approximate confidence regions using principal component methods on selected texture measures to detect and discriminate PNEU from PTB, LC and normal lungs (NL). A brief simulation study indicates that the probability ellipsoid is robust. The main result of this study is that the ellipsoids are capable of detecting pneumonia when using standard deviation of value, standard deviation of energy, minimum value, zero-crossing and homogeneity.