Single point imaging with radial acquisition and compressed sensing

Abstract

Purpose: To accelerate the Pointwise Encoding Time Reduction with Radial Acquisition (PETRA) sequence using compressed sensing while preserving the image quality for high-resolution MRI of tissue with ultra-short (Formula presented.) values. Methods: Compressed sensing was introduced in the PETRA sequence (csPETRA) to accelerate the time-consuming single point acquisition of the k-space center data. Random undersampling was applied to achieve acceleration factors up to Acc = 32. Phantom and in vivo images of the knee joint of six volunteers were measured at 3T using csPETRA sequence with Acc = 4, 8, 12, 16, 24, and 32. Images were compared against fully sampled PETRA data (Acc = 1) for structural similarity and normalized-mean-square-error. Qualitative and semi-quantitative analyses were performed to assess the effect of the acceleration on image artifacts, image quality, and delineation of anatomical structures at the knee. Results: Even at high acceleration factors of Acc = 16 no aliasing artifacts were observed, and the anatomical details were preserved compared with the fully sampled data. The normalized-mean-square-error was less than 1% for Acc = 16, in which single point imaging acquisition time was reduced from 165 to 10 s, reducing the total scan time from 7.8 to 5.2 min. Semi-quantitative analyses suggest that Acc = 16 yields comparable diagnostic quality as the fully sampled data for knee imaging at a scan time of 5.2 min. Conclusion: csPETRA allows for ultra-short (Formula presented.) imaging of the knee joint in clinically acceptable scan times while maintaining the image quality of original non-accelerated PETRA sequence.