Test-retest reliability of automated whole body and compartmental muscle volume measurements on a wide bore 3T MR system

• Published in: European radiology 2014-09, Vol.24 (9), p.2279-2291
• Main Authors: Thomas, Marianna S., Newman, David, Leinhard, Olof Dahlqvist, Kasmai, Bahman, Greenwood, Richard, Malcolm, Paul N., Karlsson, Anette, Rosander, Johannes, Borga, Magnus, Toms, Andoni P.
• Format: Article
• Language: English
• Subjects: Adult; Arthritis; Automation; Body Fluids - physiology; Body mass index; Diagnostic Radiology; Disease; Female; Healthy Volunteers; Humans; Image Processing, Computer-Assisted - methods; Imaging; Internal Medicine; Interventional Radiology; Magnetic Resonance Imaging - instrumentation; Male; Medicine; Medicine & Public Health; Middle Aged; MRI; Muscle; Muscle, Skeletal - anatomy & histology; Musculoskeletal; Neuroradiology; Organ Size; Osteoarthritis; Prospective Studies; Quality of life; Radiology; Reliablity; Reproducibility of Results; Segmentation; Ultrasound; Volume; Volunteers; Whole Body Imaging - methods
• ISSN: 0938-7994; 1432-1084; 1432-1084
• DOI: 10.1007/s00330-014-3226-6

Purpose To measure the test-retest reproducibility of an automated system for quantifying whole body and compartmental muscle volumes using wide bore 3 T MRI. Materials and methods Thirty volunteers stratified by body mass index underwent whole body 3 T MRI, two-point Dixon sequences, on two separate occasions. Water-fat separation was performed, with automated segmentation of whole body, torso, upper and lower leg volumes, and manually segmented lower leg muscle volumes. Results Mean automated total body muscle volume was 19·32 L (SD9·1) and 19·28 L (SD9·12) for first and second acquisitions (Intraclass correlation coefficient (ICC) = 1·0, 95 % level of agreement -0·32–0·2 L). ICC for all automated test-retest muscle volumes were almost perfect (0·99–1·0) with 95 % levels of agreement 1.8–6.6 % of mean volume. Automated muscle volume measurements correlate closely with manual quantification (right lower leg: manual 1·68 L (2SD0·6) compared to automated 1·64 L (2SD 0·6), left lower leg: manual 1·69 L (2SD 0·64) compared to automated 1·63 L (SD0·61), correlation coefficients for automated and manual segmentation were 0·94–0·96). Conclusion Fully automated whole body and compartmental muscle volume quantification can be achieved rapidly on a 3 T wide bore system with very low margins of error, excellent test-retest reliability and excellent correlation to manual segmentation in the lower leg. Key Points • Sarcopaenia is an important reversible complication of a number of diseases. • Manual quantification of muscle volume is time-consuming and expensive. • Muscles can be imaged using in and out of phase MRI. • Automated atlas-based segmentation can identify muscle groups. • Automated muscle volume segmentation is reproducible and can replace manual measurements.