Deep learning-accelerated image reconstruction in back pain-MRI imaging: reduction of acquisition time and improvement of image quality

• DOI: 10.1007/s11547-024-01787-x
• Publication Date: 2024-02-13T09:03:27Z
• Authors: Arne Estler; Till-Karsten Hauser; Merle Brunnée; Leonie Zerweck; Vivien Richter; Jessica Knoppik; Anja Örgel; Eva Bürkle; Sasan Darius Adib; Holger Hengel; Konstantin Nikolaou; Ulrike Ernemann; Georg Gohla
• Keywords: Artifact (error); Image noise

Abstract Introduction Low back pain is a global health issue causing disability and missed work days. Commonly used MRI scans including T1-weighted T2-weighted images provide detailed information of the spine surrounding tissues. Artificial intelligence showed promise in improving image quality simultaneously reducing scan time. This study evaluates performance deep learning (DL)-based T2 turbo spin-echo (TSE, DLR ) T1 TSE (T1 lumbar imaging regarding acquisition time, quality, artifact resistance, diagnostic confidence. Material methods retrospective monocentric included 60 patients with lower who underwent spinal between February April 2023. parameters DL reconstruction (DLR) techniques were utilized to acquire images. Two neuroradiologists independently evaluated datasets based on various using 4-point Likert scale. Results Accelerated significantly less noise artifacts, as well better sharpness, compared standard imaging. Overall confidence higher accelerated Relevant disk herniations fractures detected both conventional Both readers favored majority examinations. The examination time was cut by 61% Conclusion In conclusion, utilization learning-based resulted significant savings up imaging, while also These findings highlight potential these enhance efficiency accuracy clinical practice for pain.