Back to all papers

Paraspinal muscle radiomics from MRI alone or combined with CT for predicting outcomes after degenerative lumbar spine surgery: a scoping review.

November 16, 2026pubmed logopapers

Authors

Algahiny A,Asim I,Elnagar M,Ghorab A,Sadek M,Elmitwalli O

Affiliations (4)

  • Royal College of Surgeons in Ireland - Bahrain, Al Muharraq, Bahrain. [email protected].
  • Royal College of Surgeons in Ireland - Bahrain, Al Muharraq, Bahrain. [email protected].
  • Royal College of Surgeons in Ireland - Bahrain, Al Muharraq, Bahrain.
  • Tbilisi State Medical University, Tbilisi, Georgia.

Abstract

Paraspinal muscle degeneration is increasingly recognised as a predictor of outcomes following lumbar spine surgery. This review maps studies that used paraspinal muscle radiomics, alone or within multimodal models, to predict postoperative outcomes in lumbar degenerative disease, and appraises their methodological quality and risk of bias. This scoping review followed PRISMA-ScR. PubMed, Embase, Scopus and Web of Science were searched from inception to 10 April 2026. Eligible studies extracted radiomic or deep features from paraspinal muscles on preoperative MRI, with or without other tissues, modalities or clinical variables, and reported the discrimination of a model predicting a postoperative outcome. Data was charted with CHARMS-based items, quality was appraised with METRICS and risk of bias with PROBAST+AI. Findings were synthesised by clinical question, and MRI-only muscle models were presented separately from multimodal models. Six retrospective cohort studies were included, of which three addressed functional improvement and one study each evaluated recurrent disc herniation, radiographic adjacent segment degeneration and prolonged length of stay. Only two studies performed external validation: combined models achieved areas under the curve (AUCs) of 0.79-0.80 for functional outcome after open fusion and 0.835 for radiographic adjacent segment degeneration. Internally validated AUCs of final models ranged from 0.76 to 0.94 and depended on the resampling design. In the largest cohort, adding radiomics to clinical predictors significantly improved discrimination. In two others, incremental value was limited: radiomics added only a small gain over conventional fatty infiltration (AUC 0.922 to 0.936), and adding clinical and disc features did not improve a muscle radiomics model (0.814 vs 0.862; p = 0.083). Calibration was assessed in five studies, and none reported calibration slope or intercept. METRICS quality was good in four studies and moderate in two; under PROBAST+AI, five studies were judged at high concern or high risk of bias and one unclear. Paraspinal muscle radiomics from routine preoperative imaging is a promising way to predict outcomes after surgery for lumbar degenerative disease, using scans that patients already have. However, the evidence comes from a small number of heterogeneous retrospective studies, mostly at high risk of bias, with limited assessment of calibration and inconsistent incremental value over conventional measures. The models remain investigational, but prospective multicentre studies with rigorous pipelines, standardised outcomes, clinical comparators, external validation, calibration and decision-curve analysis could move them towards clinical use.

Topics

Journal ArticleReview

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.