The last decades have seen an increasing socioeconomic impact of obesity and obesity-related diseases. Noninvasive measures like subcutaneous and visceral adipose tissue (SAT, VAT) amounts and are also increasingly correlated with other, often clinical or metabolic findings as well as independent patient characteristics, even interventional complication rates. MRI fat quantification is common but manual processing is often laborious and time consuming while fully automatic segmentation is prone to errors. This work takes a custom-made semiautomatic MRI tool and prospectively analyzes the processing and interaction times for readers with different experience as well as patients from different BMI groups.
The last decades have seen an increasing socioeconomic impact of obesity and obesity-related diseases. Noninvasive measures like subcutaneous and visceral adipose tissue (SAT, VAT) amounts have been assessed by radiological imaging techniques and are also increasingly correlated with other, often clinical or metabolic findings as well as independent patient characteristics such as interventional complication rates.
In obesity research, segmentation of SAT and VAT areas in volumetric MRI data sets is common but manual processing is often laborious and time consuming while fully automatic segmentation is often prone to errors. This work takes a custom-made tool for semiautomatic MRI quantification and prospectively records various processing and interaction times for readers with different experience as well as images from patients over a wide range of different body mass indices (BMI).
Processing details of the custom-made graphical analysis software (under Matlab) can be found in a previous work [1]. In short, the software uses in-phase and opposed-phase Dixon images as input. The graphical user interface is shown in Fig. 1. Patients were selected from IRB-approved clinical studies involving MRI exams (1.5T Achieva XR, Philips, Best) with routine quantification of abdominal VAT and SAT volumes. Abdominal coverage–from diaphragm to pelvic floor–involved an average of 35 slices (10 mm thick, 0.5 mm gap). Data from a total of 80 patients were divided into six BMI groups: 10 normal (<26 kg/m2), 10 overweight (26–30), 20 obese (>30, OBS), 20 severely obese (>35, SVR), 10 morbidly obese (>40) and 10 “super obese” (>45, SUP).
Data were analyzed by
one experienced reader (RE, all BMI groups, n=80) and two less
experienced ones (R1 and R2, OBS+SVR only, n=40).
Automatic segmentation time (tAS), manual correction time (tMC,
adjustment of contours and thresholds) and total interaction time (tTI,
from reading of presegmented images to saving of final data) were recorded and are reported as mm:ss.