Signal-to-noise ratio assessment of muscle diffusion tensor imaging using single image set and validation by the difference image method

Abstract

Objective: Signal-to-noise ratio (SNR) assessment is essential for accurate quantification of diffusion tensor imaging (DTI) metrics and usually requires the use of a difference image method using duplicate images. We aimed to estimate the SNR of DTI of thigh muscles using a single image set without duplicate images. Methods: DTI of one thigh were acquired on a 3 T scanner from 15 healthy adults, and scans with number of signal averages (NSA) = 4 and 8 were repeatedly acquired. SNR were evaluated for six thigh muscles. For SNR calculation from a single image set, diffusion-weighted images with similar diffusion encoding directions were grouped into pairs. The difference image of each pair was high-pass filtered in k-space to yield noise images. Noise images were also calculated with a difference method using two image sets as a reference. Subjects were divided into two groups for filter optimization and validation, respectively. The coefficient of repeatability (CR) of the SNR obtained from the two methods was also evaluated separately. Results: Bland–Altman analysis comparing the single image set method and the reference showed 95% limits of agreement of −9.2 to 9.2% for the optimization group and −12.5 to 12.6% for the validation group. The SNR measurement had a CR of 21.1% using the reference method, and 13.8% using the single image set method. Conclusion: The single image method can be used for DTI SNR assessment and offers better repeatability. Advances in knowledge: SNR of skeletal muscle DTI can be assessed for any data set without duplicate images.