Comparing the number of landmark points used to create statistical shape models of the lumbar spine from MR images

Authors: Rodrigues, L., Breen, A., Knapp, K.M. and Meakin, J.R.

Journal: Radiography

Volume: 31

Issue: 5

eISSN: 1532-2831

ISSN: 1078-8174

DOI: 10.1016/j.radi.2025.103009

Abstract:

Introduction: Statistical shape models (SSMs) are used in medical imaging to quantify variations in the shape of anatomical structures across a population. This study aimed to understand if lumbar spine shape models obtained using different numbers of landmark points provide similar information. Methods: Three SSMs were created by placing landmarks points around the edges of the vertebral bodies in sagittal T2 weighted MR images of the lumbar spine from 30 women. The three SSMs differed in the number of landmark points, with four, eight, and 28 points per vertebra. Modes of variation identified by each model were compared qualitatively by three observers and the scores assigned to the modes were compared quantitatively using Pearson correlation coefficients. Results: The first five modes of variation explained approximately 80 % of the variance in shape for all three models. The first four modes described similar variations in shape (including lumbar curvature and vertebral depth) and correlated strongly (r > 0.9). The only shape descriptor that did not correlate strongly across the three models was the concavity in the edges of the vertebral body. Conclusion: SSMs created using four, eight and 28 landmark points per vertebra characterise similar features of lumbar spine shape, indicating that a small number of landmark points may be used for some applications. However, shape models created using four points are not able to characterise the concavity of the vertebral edges, limiting their use for applications where this anatomical feature is important. Implications for practice: Shape models could in the future be used opportunistically to identify and predict several musculoskeletal disorders of the spine. However, the number of landmark points needs to be chosen depending on the application of interest.

Source: Scopus

Comparing the number of landmark points used to create statistical shape models of the lumbar spine from MR images.

Authors: Rodrigues, L., Breen, A., Knapp, K.M. and Meakin, J.R.

Journal: Radiography (Lond)

Volume: 31

Issue: 5

Pages: 103009

eISSN: 1532-2831

DOI: 10.1016/j.radi.2025.103009

Abstract:

INTRODUCTION: Statistical shape models (SSMs) are used in medical imaging to quantify variations in the shape of anatomical structures across a population. This study aimed to understand if lumbar spine shape models obtained using different numbers of landmark points provide similar information. METHODS: Three SSMs were created by placing landmarks points around the edges of the vertebral bodies in sagittal T2 weighted MR images of the lumbar spine from 30 women. The three SSMs differed in the number of landmark points, with four, eight, and 28 points per vertebra. Modes of variation identified by each model were compared qualitatively by three observers and the scores assigned to the modes were compared quantitatively using Pearson correlation coefficients. RESULTS: The first five modes of variation explained approximately 80 % of the variance in shape for all three models. The first four modes described similar variations in shape (including lumbar curvature and vertebral depth) and correlated strongly (r > 0.9). The only shape descriptor that did not correlate strongly across the three models was the concavity in the edges of the vertebral body. CONCLUSION: SSMs created using four, eight and 28 landmark points per vertebra characterise similar features of lumbar spine shape, indicating that a small number of landmark points may be used for some applications. However, shape models created using four points are not able to characterise the concavity of the vertebral edges, limiting their use for applications where this anatomical feature is important. IMPLICATIONS FOR PRACTICE: Shape models could in the future be used opportunistically to identify and predict several musculoskeletal disorders of the spine. However, the number of landmark points needs to be chosen depending on the application of interest.

Source: PubMed

Comparing the number of landmark points used to create statistical shape models of the lumbar spine from MR images

Authors: Rodrigues, L., Breen, A., Knapp, K.M. and Meakin, J.R.

Journal: RADIOGRAPHY

Volume: 31

Issue: 5

eISSN: 1532-2831

ISSN: 1078-8174

DOI: 10.1016/j.radi.2025.103009

Source: Web of Science (Lite)

Comparing the number of landmark points used to create statistical shape models of the lumbar spine from MR images.

Authors: Rodrigues, L., Breen, A., Knapp, K.M. and Meakin, J.R.

Journal: Radiography (London, England : 1995)

Volume: 31

Issue: 5

Pages: 103009

eISSN: 1532-2831

ISSN: 1078-8174

DOI: 10.1016/j.radi.2025.103009

Abstract:

Introduction

Statistical shape models (SSMs) are used in medical imaging to quantify variations in the shape of anatomical structures across a population. This study aimed to understand if lumbar spine shape models obtained using different numbers of landmark points provide similar information.

Methods

Three SSMs were created by placing landmarks points around the edges of the vertebral bodies in sagittal T2 weighted MR images of the lumbar spine from 30 women. The three SSMs differed in the number of landmark points, with four, eight, and 28 points per vertebra. Modes of variation identified by each model were compared qualitatively by three observers and the scores assigned to the modes were compared quantitatively using Pearson correlation coefficients.

Results

The first five modes of variation explained approximately 80 % of the variance in shape for all three models. The first four modes described similar variations in shape (including lumbar curvature and vertebral depth) and correlated strongly (r > 0.9). The only shape descriptor that did not correlate strongly across the three models was the concavity in the edges of the vertebral body.

Conclusion

SSMs created using four, eight and 28 landmark points per vertebra characterise similar features of lumbar spine shape, indicating that a small number of landmark points may be used for some applications. However, shape models created using four points are not able to characterise the concavity of the vertebral edges, limiting their use for applications where this anatomical feature is important.

Implications for practice

Shape models could in the future be used opportunistically to identify and predict several musculoskeletal disorders of the spine. However, the number of landmark points needs to be chosen depending on the application of interest.

Source: Europe PubMed Central