Fully automatic CT-histogram-based fat estimation in dead bodies

Post-mortem body cooling is the foundation of temperature-based death time estimations (TDE) in homicide cases. Forensic science generally provides two types of p.m. body cooling models, the phenomenological and the physical models. Since both of them have to implement important individual parameters like the quantity of abdominal fat explicitly or implicitly, a more exact quantification and localization of abdominal fat is a desideratum in TDE. Particularly for the physical models, a better knowledge of the abdominal fat distribution could lead to relevant improvements in TDEs. Modern imaging methods in medicine like computed tomography (CT) are opening up the possibility to register the quantity and spatial distribution of body fat in individual cases with unprecedented precision. Since a CT-scan of an individual’s abdominal region can comprise 1000 slices as an order of magnitude, it is evident that their evaluation for body fat quantification and localization needs fully automated algorithms. The paper at hand describes the development and validation of such an algorithm called “CT-histogram-based fat estimation and quasi-segmentation” (CFES). The approach can be characterized as a weighted least squares method dealing with the gray value histogram of single CT-slices only. It does not require any anatomical a priori information nor does it perform time-consuming feature detection on the CT-images. The processing result consists in numbers quantifying the amount of abdominal body fat and of muscle-, organ-, and connective tissue. As a by-product, CFES generates a quasi-segmentation of the slices processed differentiating fat from muscle-, organ-, and connective tissue. The tool is validated on synthetic data and on CT-data of a special phantom. It was also applied on a CT-scan of a dead body, where it produced anatomically plausible results.