Purpose
Body composition and its role as a predictor of clinical outcomes is an emerging topic that has gathered significant research interest lately. For the past few decades, body composition research centred around body adipose tissues because of the growing importance of obesity as a public health concern[1]. More recently however, skeletal muscle mass has risen as another valuable measure for research[2];notably, sarcopenia gained attention as a significant factor correlating to worse outcomes in a broad variety of medical conditions and surgery[3].
Despite the increased interest...
Methods and materials
Image Acquisition
CT images for this study were extracted from a surgical audit dataset carried out locally at Aberdeen Royal Infirmary as part of the Older Persons Surgical Outcomes Collaboration (OPSOC - www.OPSOC.eu), with local institutional approval. This database contains clinical and demographic data of 75 patients over 65 years of age who underwent emergency abdominal or vascular surgery in 2016. All 75 patients in this dataset had a CT scan as part of their clinical care. CT images were extracted using Multiplanar Reconstruction (MPR)...
Results
On average, it took observer 1 (mean ± SD) 664 ± 128 seconds and observer 2 (mean ± SD) 467 ± 130 seconds to segment each image into the 6 predefined classes (P < 0.001).
Overall, there was good segmentation agreement between the two observers (Figure 6). The similarity scores do reflect however that areas which are difficult to manually segment (for example, the quadratus lumborum muscle due to its often small size, and the visceral adipose tissue due to its complexity) have weaker similarity...
Conclusion
The results of this ongoing study are encouraging. We have quantified the time burden for complete segmentation of a single CT slice for body composition analysis, demonstrated good agreement between observers of different degrees of experience and CT interpretation skills, and have proposed a semi-automated protocol for single-slice abdominal CT segmentation. However, we do acknowledge further development of this algorithm is warranted to improve its robustness, and additional work is needed to expand its functionality to fully automate the segmentation process.
Personal information and conflict of interest
J. A. Perdomo:
Nothing to disclose
B. E. Morrissey:
Nothing to disclose
G. Ramsay:
Nothing to disclose
P. K. Myint:
Nothing to disclose
R. Mitchell-Hay:
Nothing to disclose
References
[1] D. L. Duren et al., “Body Composition Methods: Comparisons and Interpretation,” Journal of Diabetes Science and Technology, vol. 2, no. 6, Nov. 2008, doi: 10.1177/193229680800200623.
[2] R. D. Boutin, L. Yao, R. J. Canter, and L. Lenchik, “Sarcopenia: Current concepts and imaging implications,” American Journal of Roentgenology, vol. 205, no. 3, 2015, doi: 10.2214/AJR.15.14635.
[3] C. Beaudart, M. Zaaria, F. Pasleau, J. Y. Reginster, and O. Bruyère, “Health outcomes of sarcopenia: A systematic review and meta-analysis,” PLoS ONE, vol. 12, no. 1, 2017, doi:...