Validating the accuracy of multispectral metal artifact suppressed diffusion-weighted imaging
John P Neri,Matthew F Koff,Kevin M Koch,Ek T Tan
DOI: https://doi.org/10.1002/mp.15925
Abstract:Background: Diffusion-weighted imaging (DWI) provides quantitative measurement of random water displacement in tissue as calculated by the apparent diffusion coefficient (ADC). While heavily utilized in stroke and oncology applications, DWI is a promising tool to map microstructural changes in musculoskeletal applications including evaluation of synovial reactions resulting from total hip arthroplasty (THA). One major challenge facing the application of DWI in THA is the significant artifacts related to the conventional echo-planar imaging (EPI) readout used. Multispectral imaging (MSI) techniques, including the multiacquisition with variable resonance image combination (MAVRIC), have been shown to effectively reduce metallic susceptibility artifacts around total joint replacements to render clinically useful images. Recently, a 2D periodically rotated overlapping parallel line with enhanced reconstruction (PROPELLER) FSE acquisition that incorporates a diffusion preparation pulse with 2D-MAVRIC has been developed to mitigate both distortion and dropout artifacts. While there have been some preliminary assessments of DWI-MAVRIC, the repeatability of DWI-MAVRIC and the effects of key parameters, such as the number of spectral bins, are unknown. Purpose: To evaluate the quantitative accuracy of DWI-MAVRIC as compared to conventional diffusion sequences. Methods: A diffusion phantom with different reference diffusivities (ADC = 113-1123 μm2 /s) was used. Scans were performed on two 1.5T MRI scanners. DWI-EPI and DWI-MAVRIC were acquired in both the axial and coronal planes. Three spatial offsets (0 cm, 10 cm left, and 10 cm right off iso-center) were used to evaluate effects of off-isocenter positioning. To assess intraday and interday repeatability, DWI-EPI and DWI-MAVRIC acquisitions were repeated on one scanner at same-day and 9-month intervals. To assess inter-scanner repeatability, DWI-EPI and DWI-MAVRIC acquisitions were compared between two scanners. ADC maps were generated with and without gradient nonlinearity correction (GNC). Linear regression, correlation, and error statistics were determined between calculated and reference ADC values. Bland-Altman plots were generated to evaluate intraday, interday, and interscanner repeatability. Results: DWI-MAVRIC had excellent correlation to reference values but at reduced linearity (r = 1.00, slope = 0.91-0.94) as compared to DWI-EPI (r = 1.00, slope = 0.99-1.01). A greater than 5% ADC bias was observed at the lowest ADC values, predominantly in the DWI-MAVRIC scans. ADC values did not vary with DWI-MAVRIC parameters. DWI-EPI acquisitions had intraday, interday, and interscanner repeatability of 3.18 μm2 /s, 19.2 μm2 /s, and 20.2 μm2 /s, respectively. DWI-MAVRIC acquisitions had inferior intraday, interday, and interscanner repeatability of 13.3 μm2 /s, 44.7 μm2 /s, 110 μm2 /s, respectively. Lower ADC errors were found at isocenter, as compared to the left and right positions. GNC reduced the absolute error by 0.31% ± 0.89%, 3.6% ± 1.4%, 0.65% ± 2.4% for the center, left, and right positions, respectively. Conclusions: DWI-MAVRIC provides good linearity with respect to reference values and good intra- and interday repeatability.