Korean Journal of Radiology

The Korean Society of Radiology (대한영상의학회)
권호 표지
DOI QR코드

DOI QR Code

Development and Validation of a Deep Learning System for Segmentation of Abdominal Muscle and Fat on Computed Tomography

  • Hyo Jung Park (Department of Radiology and Research Institute of Radiology, Asan Image Metrics, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Yongbin Shin (School of Computer Science and Engineering, Soongsil University) ;
  • Jisuk Park (Department of Radiology and Research Institute of Radiology, Asan Image Metrics, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Hyosang Kim (Department of Nephrology, Internal Medicine, Asan Medical Center, University of Ulsan College of Medicine) ;
  • In Seob Lee (Department of Surgery, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Dong-Woo Seo (Department of Emergency Medicine, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Jimi Huh (Department of Radiology, Ajou University School of Medicine and Graduate School of Medicine, Ajou University Hospital) ;
  • Tae Young Lee (Department of Radiology, Ulsan University Hospital) ;
  • TaeYong Park (School of Computer Science and Engineering, Soongsil University) ;
  • Jeongjin Lee (School of Computer Science and Engineering, Soongsil University) ;
  • Kyung Won Kim (Department of Radiology and Research Institute of Radiology, Asan Image Metrics, Asan Medical Center, University of Ulsan College of Medicine)
  • Received : 2019.06.26
  • Accepted : 2019.10.15
  • Published : 2020.01.01
Download PDF
⟨ Previous Next ⟩

Abstract

Objective: We aimed to develop and validate a deep learning system for fully automated segmentation of abdominal muscle and fat areas on computed tomography (CT) images. Materials and Methods: A fully convolutional network-based segmentation system was developed using a training dataset of 883 CT scans from 467 subjects. Axial CT images obtained at the inferior endplate level of the 3rd lumbar vertebra were used for the analysis. Manually drawn segmentation maps of the skeletal muscle, visceral fat, and subcutaneous fat were created to serve as ground truth data. The performance of the fully convolutional network-based segmentation system was evaluated using the Dice similarity coefficient and cross-sectional area error, for both a separate internal validation dataset (426 CT scans from 308 subjects) and an external validation dataset (171 CT scans from 171 subjects from two outside hospitals). Results: The mean Dice similarity coefficients for muscle, subcutaneous fat, and visceral fat were high for both the internal (0.96, 0.97, and 0.97, respectively) and external (0.97, 0.97, and 0.97, respectively) validation datasets, while the mean cross-sectional area errors for muscle, subcutaneous fat, and visceral fat were low for both internal (2.1%, 3.8%, and 1.8%, respectively) and external (2.7%, 4.6%, and 2.3%, respectively) validation datasets. Conclusion: The fully convolutional network-based segmentation system exhibited high performance and accuracy in the automatic segmentation of abdominal muscle and fat on CT images.

Keywords

Acknowledgement

This study was supported by a grant of the Korea Health Industry Development Institute (No. HI18C1216) and a grant of the National Research Foundation of Korea (No. 2017R1A2B3011475).

References

  1. Bosello O, Zamboni M. Visceral obesity and metabolic syndrome. Obes Rev 2000;1:47-56 https://doi.org/10.1046/j.1467-789x.2000.00008.x
  2. Wajchenberg BL. Subcutaneous and visceral adipose tissue: their relation to the metabolic syndrome. Endocr Rev 2000;21:697-738 https://doi.org/10.1210/edrv.21.6.0415
  3. Blauwhoff-Buskermolen S, Versteeg KS, de van der Schueren MA, den Braver NR, Berkhof J, Langius JA, et al. Loss of muscle mass during chemotherapy is predictive for poor survival of patients with metastatic colorectal cancer. J Clin Oncol 2016;34:1339-1344 https://doi.org/10.1200/JCO.2015.63.6043
  4. Kuroki LM, Mangano M, Allsworth JE, Menias CO, Massad LS, Powell MA, et al. Pre-operative assessment of muscle mass to predict surgical complications and prognosis in patients with endometrial cancer. Ann Surg Oncol 2015;22:972-979 https://doi.org/10.1245/s10434-014-4040-8
  5. Reisinger KW, van Vugt JL, Tegels JJ, Snijders C, Hulsewe KW, Hoofwijk AG, et al. Functional compromise reflected by sarcopenia, frailty, and nutritional depletion predicts adverse postoperative outcome after colorectal cancer surgery. Ann Surg 2015;261:345-352 https://doi.org/10.1097/SLA.0000000000000628
  6. Bokshan SL, Han AL, DePasse JM, Eltorai AE, Marcaccio SE, Palumbo MA, et al. Effect of sarcopenia on postoperative morbidity and mortality after thoracolumbar spine surgery. Orthopedics 2016;39:e1159-e1164 https://doi.org/10.3928/01477447-20160811-02
  7. Jones K, Gordon-Weeks A, Coleman C, Silva M. Radiologically determined sarcopenia predicts morbidity and mortality following abdominal surgery: a systematic review and metaanalysis. World J Surg 2017;41:2266-2279 https://doi.org/10.1007/s00268-017-3999-2
  8. Fukuda Y, Yamamoto K, Hirao M, Nishikawa K, Nagatsuma Y, Nakayama T, et al. Sarcopenia is associated with severe postoperative complications in elderly gastric cancer patients undergoing gastrectomy. Gastric Cancer 2016;19:986-993 https://doi.org/10.1007/s10120-015-0546-4
  9. Boutin RD, Yao L, Canter RJ, Lenchik L. Sarcopenia: current concepts and imaging implications. AJR Am J Roentgenol 2015;205:W255-W266 https://doi.org/10.2214/AJR.15.14635
  10. Jones KI, Doleman B, Scott S, Lund JN, Williams JP. Simple psoas cross-sectional area measurement is a quick and easy method to assess sarcopenia and predicts major surgical complications. Colorectal Dis 2015;17:O20-O26 https://doi.org/10.1111/codi.12805
  11. Mourtzakis M, Prado CM, Lieffers JR, Reiman T, McCargar LJ, Baracos VE. A practical and precise approach to quantification of body composition in cancer patients using computed tomography images acquired during routine care. Appl Physiol Nutr Metab 2008;33:997-1006 https://doi.org/10.1139/H08-075
  12. Shuster A, Patlas M, Pinthus JH, Mourtzakis M. The clinical importance of visceral adiposity: a critical review of methods for visceral adipose tissue analysis. Br J Radiol 2012;85:1-10 https://doi.org/10.1259/bjr/38447238
  13. McDonald AM, Swain TA, Mayhew DL, Cardan RA, Baker CB, Harris DM, et al. CT measures of bone mineral density and muscle mass can be used to predict noncancer death in men with prostate cancer. Radiology 2017;282:475-483 https://doi.org/10.1148/radiol.2016160626
  14. Shen W, Punyanitya M, Wang Z, Gallagher D, St-Onge MP, Albu J, et al. Total body skeletal muscle and adipose tissue volumes: estimation from a single abdominal cross-sectional image. J Appl Physiol (1985) 2004;97:2333-2338 https://doi.org/10.1152/japplphysiol.00744.2004
  15. Kamiya N, Zhou X, Chen H, Muramatsu C, Hara T, Yokoyama R, et al. Automated segmentation of psoas major muscle in X-ray CT images by use of a shape model: preliminary study. Radiol Phys Technol 2012;5:5-14 https://doi.org/10.1007/s12194-011-0127-0
  16. Polan DF, Brady SL, Kaufman RA. Tissue segmentation of computed tomography images using a Random Forest algorithm: a feasibility study. Phys Med Biol 2016;61:6553-6569 https://doi.org/10.1088/0031-9155/61/17/6553
  17. Chartrand G, Cheng PM, Vorontsov E, Drozdzal M, Turcotte S, Pal CJ, et al. Deep learning: a primer for radiologists. Radiographics 2017;37:2113-2131 https://doi.org/10.1148/rg.2017170077
  18. Lee H, Troschel FM, Tajmir S, Fuchs G, Mario J, Fintelmann FJ, et al. Pixel-level deep segmentation: artificial intelligence quantifies muscle on computed tomography for body morphometric analysis. J Digit Imaging 2017;30:487-498 https://doi.org/10.1007/s10278-017-9988-z
  19. Burns JE, Yao J, Chalhoub D, Chen JJ, Summers RM. A machine learning algorithm to estimate sarcopenia on abdominal CT. Acad Radiol 2019. pii: S1076-6332(19)30165-5
  20. Decazes P, Tonnelet D, Vera P, Gardin I. Anthropometer3D: automatic multi-slice segmentation software for the measurement of anthropometric parameters from CT of PET/CT. J Digit Imaging 2019;32:241-250 https://doi.org/10.1007/s10278-019-00178-3
  21. Weston AD, Korfiatis P, Kline TL, Philbrick KA, Kostandy P, Sakinis T, et al. Automated abdominal segmentation of CT scans for body composition analysis using deep learning. Radiology 2019;290:669-679 https://doi.org/10.1148/radiol.2018181432
  22. Hu P, Huo Y, Kong D, Carr JJ, Abramson RG, Hartley KG, et al. Automated characterization of body composition and frailty with clinically acquired CT. Comput Methods Clin Appl Musculoskelet Imaging (2017) 2018;10734:25-35 https://doi.org/10.1007/978-3-319-74113-0_3
  23. Prado CM, Lieffers JR, McCargar LJ, Reiman T, Sawyer MB, Martin L, et al. Prevalence and clinical implications of sarcopenic obesity in patients with solid tumours of the respiratory and gastrointestinal tracts: a population-based study. Lancet Oncol 2008;9:629-635 https://doi.org/10.1016/S1470-2045(08)70153-0
  24. Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation. IEEE Trans Pattern Anal Mach Intell 2017;39:640-651 https://doi.org/10.1109/TPAMI.2016.2572683
  25. Weisstein EW. Affine transformation. Wolfram MathWorld Web site. http://mathworld.wolfram.com/AffineTransformation.html. Published August 5, 2018. Accessed August 14, 2019
  26. Larue RT, Defraene G, De Ruysscher D, Lambin P, van Elmpt W. Quantitative radiomics studies for tissue characterization: a review of technology and methodological procedures. Br J Radiol 2017;90:20160665
  27. Perona P, Malik J. Scale-space and edge detection using anisotropic diffusion. IEEE Trans Pattern Anal Mach Intell 1990;12:629-639 https://doi.org/10.1109/34.56205
  28. Cropp RJ, Seslija P, Tso D, Thakur Y. Scanner and kVp dependence of measured CT numbers in the ACR CT phantom. J Appl Clin Med Phys 2013;14:4417
  29. Vala HJ, Baxi A. A review on Otsu image segmentation algorithm. IJARCET 2013;2:387-389
  30. Otsu N. A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern Syst 1979;9:62-66 https://doi.org/10.1109/TSMC.1979.4310076
  31. Dice LR. Measures of the amount of ecologic association between species. Ecology 1945;26:297-302 https://doi.org/10.2307/1932409
  32. Callahan LA, Supinski GS. Sepsis-induced myopathy. Crit Care Med 2009;37:S354-S367 https://doi.org/10.1097/CCM.0b013e3181b6e439
  33. Hotchkiss RS, Moldawer LL, Opal SM, Reinhart K, Turnbull IR, Vincent JL. Sepsis and septic shock. Nat Rev Dis Primers 2016;2:16045
  34. Bridge CP, Rosenthal M, Wright B, Kotecha G, Fintelmann F, Troschel F, et al. Fully-automated analysis of body composition from CT in cancer patients using convolutional neural networks. In: Stoyanov D, Taylor Z, Sarikaya D, McLeod J, Ballester MAG, Codella NCF, eds. OR 2.0 context-aware operating theaters, computer assisted robotic endoscopy, clinical imagebased procedures, and skin image analysis. Cham: Springer, 2018:204-213
  35. Popuri K, Cobzas D, Esfandiari N, Baracos V, Jagersand M. Body composition assessment in axial CT images using FEMbased automatic segmentation of skeletal muscle. IEEE Trans Med Imaging 2016;35:512-520  https://doi.org/10.1109/TMI.2015.2479252