Deep learning-based structure segmentation and intramuscular fat annotation on lumbar magnetic resonance imaging.

The models demonstrate outstanding performance, with lower error in FI measurement compared to thresholding algorithms.