Scholarlink

Purpose <p>Degenerative cervical myelopathy (DCM) is the most common cause of chronic spinal cord dysfunction globally, often presenting with subtle, nonspecific symptoms that complicate early diagnosis. While clinical examination and paraclinical investigations provide decent accuracy in diagnosis, artificial intelligence (AI), including machine learning (ML) and deep learning (DL), is being increasingly explored for its potential to enhance diagnostic accuracy in the early stages of the disease. This meta-analysis aims to assess the diagnostic performance of AI-enhanced approaches in detecting DCM.</p> Methods <p>We searched PubMed/MEDLINE, Scopus, Embase, and Web of Science (through October 2024) for human studies published after January 1, 2011, employing AI for DCM diagnosis. Two reviewers independently extracted data on participant characteristics, diagnostic modality, and the performance of the AI model. A bivariate random-effects model generated pooled estimates of sensitivity, specificity, likelihood ratios (LR+, LR−), and the area under the summary receiver operating characteristic curve (AUC). Subgroup analyses compared ML with DL, imaging with non-imaging modalities, and prospective with retrospective designs. The risk of bias was assessed using QUADAS-2, and publication bias was evaluated with Deeks’ funnel plot.</p> Results <p>Nine studies involving 1,558 participants met the inclusion criteria. Pooled sensitivity and specificity were 0.85 (95% CI, 0.75–0.91) and 0.89 (95% CI, 0.79–0.94), respectively, with an AUC of 0.93. ML-based techniques exhibited higher sensitivity than DL (0.90 vs. 0.78; <i>p</i> &lt; 0.001). Imaging-based modalities demonstrated a nonsignificant trend toward superior performance compared with non-imaging approaches. No publication bias was detected.</p> Conclusions <p>AI-enhanced diagnostic approaches, particularly ML-based modalities, demonstrate strong accuracy in diagnosing DCM. Future efforts should prioritize larger datasets, standardized protocols, and multi-institutional collaborations to refine AI’s potential role in early detection of DCM, thereby improving clinical outcomes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Degenerative cervical myelopathy diagnosis: applicability of artificial intelligence-enhanced modalities in a systematic review and meta-analysis

  • 期刊论文

  • 发表日期：

    2026年2月27日

查看全文
  • Hassan Darabi,
  • Harshit Arora,
  • Ramin Shekouhi,
  • Haydn Hoffman,
  • Reza Forghani,
  • Francis Farhadi

摘要

Purpose

Degenerative cervical myelopathy (DCM) is the most common cause of chronic spinal cord dysfunction globally, often presenting with subtle, nonspecific symptoms that complicate early diagnosis. While clinical examination and paraclinical investigations provide decent accuracy in diagnosis, artificial intelligence (AI), including machine learning (ML) and deep learning (DL), is being increasingly explored for its potential to enhance diagnostic accuracy in the early stages of the disease. This meta-analysis aims to assess the diagnostic performance of AI-enhanced approaches in detecting DCM.

Methods

We searched PubMed/MEDLINE, Scopus, Embase, and Web of Science (through October 2024) for human studies published after January 1, 2011, employing AI for DCM diagnosis. Two reviewers independently extracted data on participant characteristics, diagnostic modality, and the performance of the AI model. A bivariate random-effects model generated pooled estimates of sensitivity, specificity, likelihood ratios (LR+, LR−), and the area under the summary receiver operating characteristic curve (AUC). Subgroup analyses compared ML with DL, imaging with non-imaging modalities, and prospective with retrospective designs. The risk of bias was assessed using QUADAS-2, and publication bias was evaluated with Deeks’ funnel plot.

Results

Nine studies involving 1,558 participants met the inclusion criteria. Pooled sensitivity and specificity were 0.85 (95% CI, 0.75–0.91) and 0.89 (95% CI, 0.79–0.94), respectively, with an AUC of 0.93. ML-based techniques exhibited higher sensitivity than DL (0.90 vs. 0.78; p < 0.001). Imaging-based modalities demonstrated a nonsignificant trend toward superior performance compared with non-imaging approaches. No publication bias was detected.

Conclusions

AI-enhanced diagnostic approaches, particularly ML-based modalities, demonstrate strong accuracy in diagnosing DCM. Future efforts should prioritize larger datasets, standardized protocols, and multi-institutional collaborations to refine AI’s potential role in early detection of DCM, thereby improving clinical outcomes.