Quantitative analysis of tibial subchondral bone: Texture analysis outperforms conventional trabecular microarchitecture analysis. Issue 5 (25th November 2015)

• Record Type: Journal Article
• Authors: MacKay, James W.; Murray, Philip J.; Low, Samantha B. L.; Kasmai, Bahman; Johnson, Glyn; Donell, Simon T.; Toms, Andoni P.
• Is Part Of: Journal of magnetic resonance imaging. Volume 43:Issue 5(2016)
• Page Start: 1159
• Page End: 1170
• Publication Date: 2015-11-25
• Subjects: osteoarthritis -- magnetic resonance imaging (MRI) -- texture analysis -- subchondral bone; Magnetic resonance imaging -- Periodicals; 616
• DOI: 10.1002/jmri.25088
• Languages: English
• ISSNs: 1053-1807

Abstract

Abstract : Background: The aim of this study was to compare two different methods of quantitative assessment of tibial subchondral bone in osteoarthritis (OA): statistical texture analysis (sTA) and trabecular microarchitecture analysis (tMA). Methods: Asymptomatic controls aged 20–30 (n = 10), patients aged 40–50 with chronic knee pain but without established OA (n = 10) and patients aged 55–85 with advanced OA scheduled for knee replacement (n = 10) underwent knee MR imaging at 3 Tesla with a three‐dimensional gradient echo sequence to allow sTA and tMA. tMA and sTA features were calculated using region of interest creation in the medial (MT) and lateral (LT) tibial subchondral bone. Features were compared between groups using one‐way analysis of variance. The two most discriminating tMA and sTA features were used to construct exploratory discriminant functions to assess the ability of the two methods to classify participants. Results: No tMA features were significantly different between groups at either MT or LT. 17/20 and 11/20 sTA features were significantly different between groups at the MT/LT, respectively ( P < 0.001). Discriminant functions created using tMA features classified 12/30 participants correctly (40% accuracy; 95% confidence interval [CI], 22–58%) based on MT data and 9/30 correctly (30%, ; 95% CI, 14–46) based on LT data. Discriminant functions using sTA features classified 16/30 participants correctly (53%; 95% CI, 35–71) based on MT data and 14/30 correctly (47%; 95% CI, 29–65) based on LT data. Conclusion: sTA features showed more significant differences between the three study groups and improved classification accuracy compared with tMA features. J. Magn. Reson. Imaging 2016;43:1159–1170.