Automation of fat infiltrates segmentation from knee muscles

Abstract

This project looks specifically into Sarcopenia, a syndrome characterized by the

progressive loss of muscle mass associated with age. Sarcopenia is the central factor

in the pathophysiology of frailty.

The quantification of muscular fat infiltrates has become increasingly important, as a

number of researches have identified a relationship between the accumulation of fat

infiltrates and frailty. Therefore, the ability to quantify fats efficiently and accurately

is significant for further analysis of this field.

Segmenting the fat infiltrates from medical images is a crucial step towards their

quantification and visualization. However, there is no previous work performed on

semi-automatic or fully automatic segmentation of fat infiltrates. Mostly still depend

heavily on manual segmentation that is both tedious and time-consuming. Hence, the

aim of this project is to develop an automated segmentation program for the

extraction of fat infiltrates from MRI images of the knee muscle.

In this project, several procedures for a snake active contour program automated

successfully. Most importantly, the choosing of initialization points and the

processing of all slices associated with a particular muscle group. In addition, several

problems with the accuracy of segmentation were also identified, with the

background intensity being the most crucial factor. Deeper analysis was carried out

to obtain the optimal factor for segmentation.