R.Du Hyun RoOrthopedics & Research
Text
← The Story of Research

AI and video testing. · 2025

Can AI Measure Angles When the Spine Is Curved?

This study measures lumbar and pelvic angles automatically. Examples illustrate both small and large measurement errors.

The Spine Journal · Dong-Ho Kang, Ye-Jin Jeong, Sung Taeck Kim et al. · Du Hyun Ro

Lateral X-ray illustrating spinopelvic angles

Spinopelvic measurement example from CONNEVO SPIN product materialsEnlarge Figure

A lateral X-ray showing LL, PI, PT, SS and the reference landmarks.

CONNECTEVE · CONNEVO SPIN · Owner-supplied product brochure

Evidence & Original SourceImage use terms
01

In spine photography, you can see the pelvis together.

Measuring spinal balance involves more than how curved the back appears. Lumbar curvature and pelvic orientation help explain how the body maintains posture. This paper studies automatic measurement of lumbar lordosis, pelvic incidence, pelvic tilt, and sacral slope. It extends radiographic-measurement research from the knee to the spine and pelvis.

02

AI has been finding the bone area and the reference point in turn.

1,011 standing lateral lumbar images were used to segment bones and identify angle landmarks. The test set included 113 people, 39 with degenerative lumbar scoliosis. Against specialist references, the study assessed both repeatability and actual measurement differences.

03

I'm going to show you the right example and the wrong one.

Figure 3A shows small automated–reference differences; B shows a large error. Speed and consistency alone are insufficient. Check landmark detection in overlapping bone, indistinct femoral heads and spinal deformity. These examples clarify limitations better than average performance alone.

Results explained with study numbers

Of 113 test patients, 39 had degenerative scoliosis. LL, PI, PT and SS measurement differs from diagnostic classification. Repeat ICC 1.00 for the same image must be distinguished from agreement with specialist reference values.Enlarge Figure

Of 113 test patients, 39 had degenerative scoliosis. LL, PI, PT and SS measurement differs from diagnostic classification. Repeat ICC 1.00 for the same image must be distinguished from agreement with specialist reference values.

The Spine Journal (2025). DOI: 10.1016/j.spinee.2025.01.020 · Original explanatory chart based on reported data

04

I'm going to repeat the same answer over and over again.

Repeat-measurement agreement was 1.00, and agreement with reference measurements was high. Producing the same answer for the same input differs from measuring the true angle accurately. In a Bland–Altman plot, inspect the spread of points as well as the center line: a small mean error can coexist with large errors on some images.

05

It's not an automated study to diagnose and determine surgery.

This study assessed angle-measurement performance, not who should have surgery or whose postoperative pain will improve. Testing with other institutions and equipment is needed. Clinicians should check automated results against the original image before combining them with symptoms and examination.

Look at the study numbers again.

1,011Lateral radiographs analyzed
113Test group
39Participants with scoliosis in the test group
RESEARCH FINDINGThe video.

Composition of the study sample

Lateral radiographs analyzed
1,011
0581.321,162.65
Counts of data included in the study. These do not show treatment effects or an individual’s risk.

Common Questions

Can we apply these results directly to individual therapies?

This study assessed angle-measurement performance, not who should have surgery or whose postoperative pain will improve. Testing with other institutions and equipment is needed. Clinicians should check automated results against the original image before combining them with symptoms and examination.

Methods and Numbers in Detail

Automated measurement of pelvic parameters using convolutional neural network in complex spinal deformities: overcoming challenges in coronal deformity cases

Participants and Methods

1,011 standing lateral lumbar radiographs. Of 113 test patients, 39 (34.5%) had degenerative lumbar scoliosis.

U-Net segmentation and Keypoint R-CNN were combined to calculate LL, PI, PT, and SS. Results were compared with specialist reference values and two readers, using Bland–Altman analysis, ICC, and 5-fold cross-validation.

Results Reported in the Paper

  1. Repeat-measurement ICC was 1.00 for every angle. Agreement with reference measurements and cross-validation ICCs were 0.96–0.99.
  2. Figure 2 reports mean test-set errors of PI 0.23°, LL 0.05°, SS 0.03°, and PT 0.20°. The range of individual errors differed by angle.
  3. Errors and agreement in the degenerative-scoliosis subgroup differed from the general group. Figure 3 shows both an accurate example and an example with a large error.

Scope of Interpretation

This retrospective single-center study needs external validation using other equipment and institutions. A repeatability ICC of 1.00 means that the same input produces the same output; it does not mean that every image has an accurate angle measurement. The 39 participants with scoliosis in the test set had a mean Cobb angle of 14.8°. The results cannot be generalized to all severe deformities or postoperative images.

Keywords

Abstract · Original text

Dong-Ho Kang, Ye-Jin Jeong, Sung Taeck Kim, Younguk Kim, Bong-Soon Chang, Hyoungmin Kim, Sam Yeol Chang, Du Hyun Ro. Automated measurement of pelvic parameters using convolutional neural network in complex spinal deformities: overcoming challenges in coronal deformity cases. The spine journal : official journal of the North American Spine Society. 2025;25(8):1688-1697. doi:10.1016/j.spinee.2025.01.020.

BACKGROUND CONTEXT

Accurate and consistent measurement of sagittal alignment is challenging, particularly in patients with severe coronal deformities, including degenerative lumbar scoliosis (DLS).

PURPOSE

This study aimed to develop and validate an artificial intelligence (AI)-based system for automating the measurement of key sagittal parameters, including lumbar lordosis, pelvic incidence, pelvic tilt, and sacral slope, with a focus on its applicability across a wide range of deformities, including severe coronal deformities, such as DLS.

DESIGN

Retrospective observational study.

PATIENT SAMPLE

A total of 1,011 standing lumbar lateral radiographs, including DLS.

OUTCOME MEASURE

Interclass and intraclass correlation coefficients (CC), and Bland-Altman plots.

METHODS

The model utilizes a deep-learning framework, incorporating a U-Net for segmentation and a Keypoint Region-based Convolutional Neural Network for keypoint detection. The ground truth masks were annotated by an experienced orthopedic specialist. The performance of the model was evaluated against ground truth measurements and assessments from two expert raters using interclass and intraclass CC, and Bland-Altman plots.

RESULTS

In the test set of 113 patients, 39 (34.5%) had DLS, with a mean Cobb's angle of 14.8°±4.4°. The AI model achieved an intraclass CC of 1.00 across all parameters, indicating perfect consistency. Interclass CCs comparing the AI model to ground truth ranged from 0.96 to 0.99, outperforming experienced orthopedic surgeons. Bland-Altman analysis revealed no significant systemic bias, with most differences falling within clinically acceptable ranges. A 5-fold cross-validation further demonstrated robust performance, with interclass CCs ranging from 0.96 to 0.99 across diverse subsets.

CONCLUSION

This AI-based system offers a reliable and efficient automated measurement of sagittal parameters in spinal deformities, including severe coronal deformities. The superior performance of the model compared with that of expert raters highlights its potential for clinical applications.

Read abstract on PubMed

Related Reading

AI and video testing.Can AI help identify fractures?AI and video testing.What Does AI Detect on Full-Length Leg X-rays to Measure Alignment?
Arthritis BasicsFind Related Articles