Accuracy and precision in the detection of articular cartilage lesions using magnetic resonance imaging at 1.5 tesla in an in vitro study with orthopedic and histopathologic correlation

Abstract

Magnetic resonance (MR) sequences for cartilage visualization have been the target of numerous studies, and the optimal sequence for cartilage imaging remains a matter of debate in the literature. To compare MR findings with different MR sequences for the detection of cartilage lesions in fresh deep-frozen human cadaveric patellae in an in vitro setting. Ten cadaveric patellae were imaged on a 1.5T MR scanner with a 2x2 channel carotid sandwich coil and a conventional knee coil, and compared with orthopedic findings and gold-standard histopathology. MR sequences were: a) fat-saturated (FS) proton density-weighted (PDw) turbo spin-echo (TSE) sequence (TR/TE 4000/39 ms); b) T2-weighted (T2w) double-echo steady-state (DESS) 3D water-excitation (we) sequence (TR/TE 17/4.7 ms); c) 3D-PDw-SPACE (sampling perfection with application-optimized contrasts using different flip-angle evolutions)-we sequence (TR/TE 1800/19 ms). Accuracy, Kendall's tau-b correlation, and weighted kappa coefficients were calculated. Accuracy for cartilage lesion detection with the FS PDw-TSE sequence and the carotid coil was 78.3%, and with the knee coil 73.9%. For the T2wDESS-3D-we sequence, the corresponding values were 69.5% and 65.2%, and for the 3D-PDw-SPACE-we sequence 65.2% and 60.8%, respectively. Kendall's tau-b correlation ranged between 0.508 for the 3D-PDw-SPACE-we sequence (knee coil) and 0.720 for the FS PDw-TSE sequence (carotid and knee coil). Weighted kappa coefficient was lowest for the 3D-PDw-SPACE-we sequence (knee coil) at 0.607, and highest for the carotid coil and FS PDw-TSE sequence at 0.779. The evaluated FS PDw-TSE sequences are superior in comparison to the T2wDESS-3D-we and 3D-PDw-SPACE-we sequences in the in vitro setting for the detection of cartilage lesions, and are comparable to results reported in the literature.