Statistical shape modelling of the human lumbar spine: a scoping review characterising the research landscape and potential clinical applications.

OBJECTIVE

A systematic scoping review was conducted to characterise the current research landscape of statistical shape modelling of the lumbar spine, focusing on the proposed clinical applications of the method.

METHODS

MEDLINE, Embase, IEEE Xplore, and Scopus were searched for sources published before 17 June 2026. Sources were screened against predefined inclusion and exclusion criteria and data were extracted using a structured form in Covidence.

RESULTS

2,930 records were identified for review. Following duplicate removal and screening, 85 studies remained for data extraction. Most studies were exploratory (57%) or descriptive (14%) in nature. The vertebral body was the most frequently isolated region, with L5 often excluded due to its anatomical distinctiveness. Reporting of population demographics and dataset composition was highly inconsistent, highlighting substantial heterogeneity in both methodology and population selection.

CONCLUSION

The use of lumbar spine statistical shape models (SSMs) has been predominantly exploratory or descriptive. Whilst prognostic, diagnostic, and therapeutic applications have been proposed, none of the reviewed studies demonstrated a direct change to clinical practice or impact on patient care. No consensus exists on optimal approaches to multi-level modelling, vertebral level pooling, or anatomical boundary definition, and training datasets are predominantly drawn from Western, female-skewed populations. In the short term, SSMs are well suited to exploratory analyses of clinical populations, revealing patterns not captured by conventional measures. Progress will benefit from training data aligned to defined clinical questions, adequate sample sizes and demographic breadth, and consistent reporting of validation metrics such as compactness, generalisability and specificity. This will enable development towards a technology that can lead to meaningful clinical translation.