Artificial intelligence-aided CT segmentation for body composition analysis: a validation study

Abstract

Background Body composition is associated with survival outcome in oncological patients, but it is not routinely calculated. Manual segmentation of subcutaneous adipose tissue (SAT) and muscle is time-consuming and therefore limited to a single CT slice. Our goal was to develop an artificial-intelligence (AI)-based method for automated quantification of three-dimensional SAT and muscle volumes from CT images. Methods Ethical approvals from Gothenburg and Lund Universities were obtained. Convolutional neural networks were trained to segment SAT and muscle using manual segmentations on CT images from a training group of 50 patients. The method was applied to a separate test group of 74 cancer patients, who had two CT studies each with a median interval between the studies of 3 days. Manual segmentations in a single CT slice were used for comparison. The accuracy was measured as overlap between the automated and manual segmentations. Results The accuracy of the AI method was 0.96 for SAT and 0.94 for muscle. The average differences in volumes were significantly lower than the corresponding differences in areas in a single CT slice: 1.8% versus 5.0% (p < 0.001) for SAT and 1.9% versus 3.9% (p < 0.001) for muscle. The 95% confidence intervals for predicted volumes in an individual subject from the corresponding single CT slice areas were in the order of ± 20%. Conclusions The AI-based tool for quantification of SAT and muscle volumes showed high accuracy and reproducibility and provided a body composition analysis that is more relevant than manual analysis of a single CT slice.

Automated artificial intelligence-based analysis of skeletal muscle volume predicts overall survival after cystectomy for urinary bladder cancer

“…The main difference between these studies and the present one is the analysis of muscle volume compared to skeletal muscle area in a single axial cross-section at the level of the L3 vertebra. We have previously shown that these different measurements correlate, but that muscle volume has a lower variance and may therefore be a better and more reliable measure of sarcopenia [8]. The present study builds on that by showing that only very few CT studies could not be correctly analysed and that the association with OS holds true.…”

“…The AI-based tool consisted of a single convolutional neural network (CNN), that had previously been trained to segment muscle and fat [8]. In addition, the CNN had also previously been trained to detect and segment the sacrum and coccyx.…”

“…In addition, the CNN had also previously been trained to detect and segment the sacrum and coccyx. These skeletal parts were used for determining the levels for muscle segmentation, which is a difference from [8] which used the 11th thoracic vertebra (Th11). The reason for this is twofold.…”

“…Artificial intelligence (AI)-based image analysis models have been developed which can accurately segment muscle and adipose tissue, both in single cross-sections [6] and in full abdominal CT scans [7]. We have previously reported on an AI-based, fully automated image analysis of abdominal CT images, which identifies skeletal muscle and subcutaneous fat and calculates the respective volumes [8]. These volumes have been shown to correlate with the respective area of a single cross-section slice, but with a lower variance and could therefore potentially provide a more reliable measure of sarcopenia.…”

Computed tomography-based multiple body composition parameters predict outcomes in Crohn’s disease

“…A detailed method of using ImageJ for the quantification of body compositions has been described previously [20]. Semi-automated segmentation was used during these manual courses by using tissuespecific attenuation thresholds such as skeletal muscle (− 29 to 150 HU) and adipose tissue (− 190 to − 30 HU) [21,22]. An example of segmented each type of tissue was shown in Additional file 1: Fig.…”