Quantitative MRI techniques for measuring fatty infiltration in the lumbar paraspinal muscles: a systematic review

Abstract Background Fatty infiltration of the lumbar paraspinal muscles has been associated with spine degeneration and is widely studied in relation to spinal disorders. Assessment of muscle fat content using magnetic resonance imaging (MRI) provides a clinically practical and non-invasive method for these studies. However, considerable methodological heterogeneity exists in the MRI-based approaches used to quantify muscle fat, and a comprehensive overview of these approaches is lacking. Therefore, this systematic review aims to summarize the MRI sequences, muscle segmentation and fat quantification methods used in research studies investigating lumbar spinal musculature and to highlight current applications of machine-learning methods in this context. Methods A systematic search was conducted on PubMed, Web of Science, and Scopus in October 2025. All retrieved publications were screened according to predefined inclusion and exclusion criteria. Relevant information regarding imaging sequences, fat quantification methods, method reliabilities and participant characteristics was extracted from the included studies. Results Of the 280 MRI sequence applications identified across the 274 included publications, 209 (75%) used either T1-weighted imaging (T1WI) or T2-weighted imaging (T2WI), while the remaining sequence applications were based on chemical-shift (CS) MRI techniques. Manual segmentation was the predominant approach for outlining the muscles of interest. Only a small number of studies used machine learning approaches for segmentation and fat quantification, with the U-Net architecture being the most common. For T1WI/T2WI images, several methods were used to quantify fatty infiltration within the muscles, including different thresholding approaches, signal-intensity measures, and manually outlining fat tissue. Reported reliabilities for fat quantification methods using T1WI/T2WI images varied across studies and were reported inconsistently. CS techniques most frequently used the 6-point Dixon method, followed by 2-point and 3-point Dixon methods for determining fatty infiltration. Studies using T1WI/T2WI images mainly investigated populations with degenerative lumbar pathologies, while the cohort who underwent CS MRI included both healthy and diseased participants. Conclusions T2WI MRI was the most commonly used imaging technique for assessing fatty infiltration of the lumbar paraspinal muscles, although the applied fat quantification methods vary and lack validation. Therefore, standardized and validated T2WI protocols are urgently needed to improve clinical utility and comparability across studies. In contrast, CS-based techniques provide a quantitative approach based on the difference in resonance frequencies between water and fat proton signals but remain less commonly used in clinical research. Future studies should focus on developing more reliable automated approaches for muscle segmentation to reduce manual workload.