A Method to Estimate Cause of Death by Texture Analysis in Organ Regions of Postmortem CT Images

Abstract

In this paper, the authors propose a method to estimate cause of death for postmortem CT images by performing texture analysis for computer-aided diagnosis (CAD) in autopsy imaging (Ai). Twenty-eight postmortem CT images were used in this study. Texture features such as Haralick's features are calculated in liver regions and lung regions extracted from postmortem CT images. The number of the calculated features is 7206 for each image. The number of features is reduced to estimate cause of death by using Correlation based Feature Selection (CFS). As a result of the experiments, the cause of death was classified with low classification error.