Effects of motion and b‐value on apparent temperature measurement by diffusion‐based thermometry MRI: eye vitreous study

Purpose To make noninvasive measurements of temperature in the posterior chamber (vitreous) of the eye using diffusion‐based thermometry (DBT) magnetic resonance imaging (MRI) and to explain variability in these measurements due to choice of b‐value and the effects of motion. Methods Phantom studies of human vitreous and distilled water were performed using b‐values from 0 to 1500 s/mm 2 to determine the liquid‐specific calibration factor for vitreous as well as to determine the temperature offsets due to sampling the diffusion curve using three higher routine clinical b‐values (b = 0, 500, 1000 s/mm 2 ) or four lower b‐values (b = 0, 200, 400, 600 s/mm 2 ), thought to be optimized for fluids. Retrospective ROI‐based measurements of apparent diffusion coefficient on single slices as well as multi‐slice histograms of the eyes were made in six patients with peri‐orbital cellulitis and 11 age‐matched controls, to assess for temperature changes in the presence of peri‐orbital inflammation. A prospective study of ten repeated measurements of eye temperature using both high and lower b‐value sampling was performed in ten asymptomatic volunteers to determine the reproducibility of eye temperature measurements in‐vivo as well as to estimate vitreous temperature in the absence of motion. Results The diffusion coefficient of vitreous (2,088 ± 13 × 10 −6 mm 2 /s) was significantly lower (−1.9%, P  < 0.001) compared to distilled water (2,128 ± 12 × 10 −6 mm 2 /s). The calibration factor for temperature measurements of vitreous using DBT is +0.74 ± 0.06°C. Temperature offsets were smaller (<−0.2°C, P  < 0.01) when using larger routine clinical b‐values to estimate the diffusion coefficient compared to using a series of lower b‐values (<−1.0°C, P  < 0.001). Two‐dimensional single‐slice ROI‐based measurement showed significant temperature differences (ΔT I‐C  = 2.5 ± 1.2°C, P  < 0.001) between the eyes of patient with peri‐orbital cellulitis, higher on the side of inflammation. There was no significant difference in eye temperature when using the 3D histogram (which is likely due to motion averaging as significant slice‐to‐slice variation was present). However, significant differences in the 3D temperature histograms between the two eyes was observed in one out of six patients. Prospective eye temperature measurements in healthy volunteers showed significant intra‐ and inter‐subject variability (33.8–41.6°C), which was caused by eye motion. This resulted in +2.4°C cohort‐wide elevation in temperature when three b‐values were used and +4.7°C when four b‐values were used. Using a pattern of elevated temperature at the periphery of the eye to detect motion, eye temperature is the absence of motion was estimated to be 34.5 ± 0.4°C with three higher b‐values and 34.6 ± 1.9°C with four lower b‐values; this temperature corresponds with prior mathematical simulations of eye temperature as well as boundary conditions. Conclusions Globe vitreous temperature has been measured noninvasively using DBT MRI. Using routine clinical b‐values of b = 0, 500 and 1000 s/mm 2 produces acceptable (<−0.2°C) temperature offsets. Although DBT measurements are highly susceptible to motion, methods such as temperature differences or regression can be used to reduce or eliminate the effects of motion. Using a single clinical diffusion‐weighted MRI, globe temperature difference of 1.6°C is pathological. Using a series of ten measurements, globe temperature differences larger than 0.6°C are abnormal. This study suggests CSF flow likely artifactually increases core brain temperature measured by DBT MRI.