Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jul 10, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 2025

Automated scoliosis X-ray cobb angle classification: a deep learning approach with RadImageNet.

Jennifer Yu1, Yash Lahoti2, Hamza Ahmed2

  • 1Icahn School of Medicine at Mount Sinai, New York, USA. jennifer.yu@icahn.mssm.edu.

European Spine Journal : Official Publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society
|July 9, 2026
PubMed
Summary

Related Concept Videos

Computed Tomography01:10

Computed Tomography

Tomography refers to imaging by sections. Computed tomography (CT) is a non-invasive imaging technique that uses computers to analyze several cross-sectional X-rays to reveal minute details about structures in the body.
The technique was invented in the 1970s and is based on the principle that as X-rays pass through the body, they are absorbed or reflected at different levels. In the technique, a patient lies on a motorized platform while a computerized axial tomography (CAT) scanner rotates...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Atrial Fibrillation/Flutter After Transcatheter Device vs. Surgical Closure of Atrial Septal Defect in a Korean Nationwide Cohort.

Korean circulation journal·2026
Same author

Effectiveness of a Teaching Session on Interpreting Common Orthopaedic Radiographs at a Major Trauma Centre.

Cureus·2026
Same author

Reducing Delayed Discharges After Neck of Femur Fracture Surgery in a Major Trauma Centre: A Retrospective Quality Improvement Study.

Cureus·2026
Same author

ONT-only genome assembly of a Korean male individual using a semen sample.

Genes & genomics·2026
Same author

Management of Femoral Head Fractures: A Systematic Review.

Cureus·2026
Same author

Premature ventricular contractions burden and long-term ventricular remodelling in patients without structural heart disease.

Heart (British Cardiac Society)·2026

A deep learning model using RadImageNet demonstrated 76.2% accuracy in classifying adult scoliosis severity from X-rays. While promising for adjunctive grading, it requires further validation for clinical use.

Area of Science:

  • Radiology
  • Artificial Intelligence
  • Medical Imaging

Background:

  • Biplanar full-body X-ray imaging is essential for scoliosis assessment and treatment guidance.
  • Manual measurement of coronal curvature is time-consuming and prone to variability.
  • Deep learning offers potential for automated, standardized scoliosis analysis.

Purpose of the Study:

  • To evaluate the efficacy of RadImageNet transfer learning for classifying adult scoliosis severity.
  • To assess the performance of an AI model in categorizing scoliosis as none, mild, or severe.

Main Methods:

  • A retrospective cohort of 816 adult scoliosis radiographs was used, with 80/20 train/test split.
  • A ResNet-50 convolutional neural network, pretrained on RadImageNet, was fine-tuned for classification.
Keywords:
AIArtificial IntelligenceCobb AngleEOSRadImageNetSpine

Related Experiment Videos

Last Updated: Jul 10, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 2025

  • Data augmentation and Smooth Grad-CAM++ were employed for enhanced generalizability and interpretability.

    • Main Results:

    • The fine-tuned model achieved 76.2% overall accuracy and an AUROC of 0.853 for severe scoliosis.
    • Performance varied by class, with the highest precision (0.79) and recall (0.90) for mild scoliosis.
    • Class activation maps indicated thoracic and lumbar regions were key to predictions, with misclassifications mainly between adjacent severity levels.

    Conclusions:

    • Fine-tuning RadImageNet shows potential for adjunctive scoliosis severity categorization.
    • The model does not directly measure Cobb angle or surgical parameters.
    • External validation is necessary before clinical implementation for screening.