跳到正文
原文
Nature 健康护理主题·· 15 小时前AI 评分45

多模态医学影像的自动脊柱异常检测:深度学习范式

Automated spinal abnormality detection using multi-modal medical imaging: a deep learning paradigm

AI 导读

该研究提出了一种基于深度学习的综合框架,用于多模态医学影像中的腰椎异常自动诊断,包括MRI和X光图像。该系统通过先进的预处理、改进的U-Net分割架构和混合特征提取,实现了95.2%的准确率、93.8%的精确率、94.5%的召回率和94.1%的F1分数,显著优于标准CNN基线和非分割方法。该方法在MRI和X光模态上表现优异,证明了结合显式脊柱分割和混合深度学习分类在多模态异常检测中的有效性。

正文

We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply.

Abstract

Spinal disorders, like herniated discs, spinal stenosis and spondylolisthesis, affect over 540 million people worldwide. However, these disorders are determined based on a manual interpretation of MRI and X-ray images. This is challenging because of the inter-observer variability of (10-30%) and the laborious process involved in the analysis undertaken. This investigation exhibits a consolidated Deep Learning framework that automates the diagnosis of lumbar spine anomalies in medical images taken from multiple modalities, overcoming the limitations of manual diagnosis and single modality approaches. We collected a dataset of 2847 images (1542 MRI and 1305 X-ray) of anonymous patients from four medical centres comprising of four classes, namely Normal, Herniated Disc, Spinal Stenosis, and Spondylolisthesis. The system consists of (1) advanced preprocessing (Gaussian & Bilateral filtering followed by CLAHE), (2) modified U-Net architecture for accurate spinal segmentation with Dice=0.91, (3) Hybrid feature extraction with combination of custom design CNN Layers with transfer learning (ResNet50, VGG19) by feature level fusion, (4) Multi-class Classification. Segmentation is interpretable through attention maps. The proposed method produced 95.2% accuracy (95% CI: 94.1-96.3%), 93.8% Precision, 94.5% Recall, and 94.1% F1 Score with extensive augmentation and 5-fold cross-validation. Our method outperforms standard CNN baselines (+3.7%) and non-segmented methods (+4.1%). The framework consistently performed well on both MRI (96.1%) and X-ray (94.3%) modalities, demonstrating that explicit spinal segmentation merged with hybrid deep learning classification can robustly detect abnormality across multi-modality. The system detects abnormalities in different imaging modalities robustly. Performance is encouraging on MRI (96.1%) and X-ray (94.3%) modalities.

Funding

Open access funding provided by Manipal University Jaipur.

Author information

Author notes

  1. Shivendra Dubey, Vasima Khan, Durgesh Nandan and Sakshi Dubey contributed equally to this work.

Authors and Affiliations

  1. CSE-Artificial Intelligence and Data Science, Sagar Institute of Science and Technology, Gandhi Nagar, Bhopal, Madhya Pradesh, 462036, India

    Ruchi Jain & Vasima Khan

  2. Department of Artificial Intelligence and Machine Learning, Manipal University Jaipur, Jaipur, Rajasthan, 303007, India

    Shivendra Dubey

  3. School of CS and AI, SR University, Warangal, Telangana, 506371, India

    Durgesh Nandan

  4. Department of Computer Science and Engineering, Manipal University Jaipur, Jaipur, Rajasthan, 303007, India

    Sakshi Dubey

Authors

  1. Ruchi Jain
  2. Shivendra Dubey
  3. Vasima Khan
  4. Durgesh Nandan
  5. Sakshi Dubey

Corresponding authors

Correspondence to Shivendra Dubey or Vasima Khan.

Ethics declarations

Ethics declaration

All methods were carried out in accordance with relevant guidelines and regulations. The study was approved by the Institutional Research & Development Committee of Sagar Institute of Science & Technology. The data taken from medical centers were fully anonymized before receiving, and hence the requirement for informed consent was waived by the Institutional Research & Development Committee of Sagar Institute of Science & Technology and medical centers.

Competing interests

The authors declare no competing Interests.

Additional information

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Jain, R., Dubey, S., Khan, V. et al. Automated spinal abnormality detection using multi-modal medical imaging: a deep learning paradigm. Sci Rep (2026). https://doi.org/10.1038/s41598-026-70849-z

Download citation

  • Received: 15 April 2026

  • Accepted: 04 September 2026

  • Published: 20 November 2026

  • DOI: https://doi.org/10.1038/s41598-026-70849-z

Keywords

来源:Nature 健康护理主题 · nature.com