AI-Generated Synthetic STIR of the Lumbar Spine from T1 and T2 MRI Sequences Trained with Open-Source Algorithms.

Abstract

Background and purpose: Lumbar spine MRIs can be time consuming, stressful for patients, and costly to acquire. In this work, we train and evaluate open-source generative adversarial network (GAN) to create synthetic lumbar spine MRI STIR volumes from T1 and T2 sequences, providing a proof-of-concept that could allow for faster MRI examinations.

Materials and methods: A total of 1817 MRI examinations with sagittal T1, T2, and STIR sequences were accumulated and randomly divided into training, validation, and test sets. A GAN was trained to create synthetic STIR volumes by using the T1 and T2 volumes as inputs, optimized with the validation set, and then applied to the test set. Acquired and synthetic test set volumes were independently evaluated in a blinded, randomized fashion by 3 radiologists specializing in musculoskeletal imaging and neuroradiology. Readers assessed image quality, motion artifacts, perceived likelihood of the volume being acquired or synthetic, and the presence of 7 pathologies.

Results: The optimal model leveraged a customized loss function that accentuated foreground pixels, achieving a structural similarity imaging metric of 0.842, mean absolute error of 0.028, and peak signal-to-noise ratio of 26.367. Radiologists could distinguish synthetic from acquired volumes; however, the synthetic volumes were of equal or better quality in 77% of test patients and demonstrated equivalent or decreased motion artifacts in 78% of test patients. For common pathologies, the synthetic volumes had high positive predictive value (75%-100%) but lower sensitivity (0%-67%).

Conclusions: This work links objective computer vision performance metrics and subject clinical evaluation of synthetic spine MRIs by using open-source and reproducible methodologies. High-quality synthetic volumes are generated, reproducing many important pathologies and demonstrating a potential means for expediting imaging protocols.