Magnetic resonance imaging (MRI) is a state-of-the-art diagnostic imaging modality which beneficially features excellent soft tissue contrast and lacks use of ionizing radiation. In clinical practice, MRI generates contrast-weighted images based on the intrinsic tissue parameters of the subject, which are then interpreted by a radiologist. However, current qualitative MRI images can have limited sensitivity to subtle changes in tissue parameters that can provide useful information for various biomedical applications, such as early detection of tumor growth and neurodegenerative diseases. While there is a promise of using quantitative MRI approach, lengthy imaging times have delayed its translation to clinical use. An emerging quantitative MRI technique known as MR Fingerprinting has demonstrated a great potential of transforming quantitative MRI by enabling simultaneous acquisition of multiple MR tissue parameter maps (e.g., T1, T2, and spin density) in a single imaging experiment. Its use of randomized parameter encoding has led to promising initial results. However, it has been shown that this scheme has sub-optimal signal-to-noise ratio (SNR) efficiency. To improve SNR, MR Fingerprinting acquisition schemes was posed as an optimal experimental design (OED) problem. An OED framework solution enabled better estimation and SNR performance. And yet, the MR Fingerprinting-OED problem has been computationally expensive to solve, which again impairs its practical and clinical utility.
Researchers at UT Austin have developed a computationally efficient approach to solve the MR Fingerprinting-OED problem. This invention exploits an early observation that the optimized data acquisition parameters appear to be highly structured. The technology efficiently incorporates this prior knowledge by representing the data acquisition parameter sequences using a class of piecewise smooth mathematical functions known as B-splines. This representation reduces the search space for the experimental design, significantly improving the computational efficiency of the OED problem.
The patent-pending software solves the commercially relevant problem of slow quantitative MRI acquisition without compromise in performance, while achieving a similar or slightly better SNR of the MR Fingerprinting imaging experiments as the state-of-the-art approaches. By improving the computational efficiency of designing MR Fingerprinting experiments by 100x, it allows MR Fingerprinting experiments to be completed in less than a minute. This enhanced performance opens up the utility of the software for a variety of quantitative MRI applications in the diagnostic, clinical, and medical fields.
MR fingerprinting using b-splines presents an opportunity to companies in the quantitative MRI space, looking for software to increase efficiency in MR acquisition times and better SNR. The University of Texas at Austin is looking for an industry partner to license this technology.