T₂ measurement in articular cartilage: Impact of the fitting method on accuracy and precision at low SNR

Abstract

T 2 relaxation time is a promising MRI parameter for the detection of cartilage degeneration in osteoarthritis. However, the accuracy and precision of the measured T 2 may be substantially impaired by the low signal-to-noise ratio of images available from clinical examinations. The purpose of this work was to assess the accuracy and precision of the traditional fit methods (linear least-squares regression and nonlinear fit to an exponential) and two new noise-corrected fit methods: fit to a noisecorrected exponential and fit of the noise-corrected squared signal intensity to an exponential. Accuracy and precision have been analyzed in simulations, in phantom measurements, and in seven repetitive acquisitions of the patellar cartilage in six healthy volunteers. Traditional fit methods lead to a poor accuracy for low T 2 , with overestimations of the exact T 2 up to 500%. The noise-corrected fit methods demonstrate a very good accuracy for all T 2 values and signal-to-noise ratio. Even more, the fit to a noise-corrected exponential results in precisions comparable to the best achievable precisions (Cramér-Rao lower bound). For in vivo images, the traditional fit methods considerably overestimate T 2 near the bone-cartilage interface. Therefore, using an adequate fit method may substantially improve the sensitivity of T 2 to detect pathology in cartilage and change in T 2 follow-up examinations. Magn Reson Med 63:181-193, 2010.

Ultrashort–Echo Time MR Imaging of the Patella with Bicomponent Analysis: Correlation with Histopathologic and Polarized Light Microscopic Findings

“…A P value of less than .01 was considered to indicate a significant difference. (25,26). Because only three parameters, namely short T2*, long T2*, and the short or long T2* fraction, were fitted in this bicomponent model, robust fitting could be achieved with clinically achievable signal-to-noise ratios (16).…”

Ultrashort echo time imaging with bicomponent analysis

“…To allow noise corrected bicomponent fitting on a voxel by voxel basis, we adopted the following model (27,28,31):…”

“…2, and I n is the modified Bessel function of the first kind of nth order. A recent article by Raya et al (31) showed that this noise corrected model is the most accurate and precise method of calculating T 2 , especially for low SNR images. After fitting was finished, goodness of fit statistics including the R-squared value, mean squared error, root mean squared error, and standard error were computed.…”

Probing articular cartilage damage and disease by quantitative magnetic resonance imaging

“…There may also be little distinction in T 2 values within the spectrum of pre-radiographic to radiographic cartilage damage [34,90]. In addition to being affected by the angle at which cartilage is positioned with respect to the main magnetic field [55], T 2 values are sensitive to the type of radiofrequency coil [91], the choice of pulse sequence and sequence parameters [92], the joint position during imaging [93] and the type of equation used to fit data [94].…”