JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

AI-Derived Blood Biomarkers for Ovarian Cancer Diagnosis: Systematic Review and Meta-Analysis

  • 0Department of Clinical Epidemiology, Shengjing Hospital of China Medical University, ShenYang, China.
Journal of Medical Internet Research +

|

March 24, 2025

|

March 24, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Artificial intelligence (AI) shows promise in diagnosing ovarian cancer (OC) using blood biomarkers, with high pooled sensitivity and specificity. Further research should focus on external validation and deep learning for improved diagnostic accuracy.

Area Of Science

  • Biomarker Discovery
  • Artificial Intelligence in Medicine
  • Oncology Diagnostics

Background

  • Artificial intelligence (AI) shows potential for identifying noninvasive blood biomarkers.
  • The diagnostic accuracy of AI-derived biomarkers for ovarian cancer (OC) requires further investigation.

Purpose Of The Study

  • To evaluate the research quality and diagnostic validity of AI-based blood biomarkers for ovarian cancer.
  • To perform a meta-analysis of diagnostic accuracy studies.

Main Methods

  • Systematic literature search across multiple databases (MEDLINE, Embase, IEEE Xplore, PubMed, Web of Science, Cochrane Library).
  • Quality assessment using the Quality Assessment of Diagnostic Accuracy Studies-AI tool.
  • Meta-analysis of pooled sensitivity, specificity, and area under the curve (AUC) using a bivariate model.

Main Results

  • Included 40 studies, with most (78%) having a low risk of bias.
  • Overall pooled diagnostic accuracy: sensitivity 85%, specificity 91%, AUC 0.95.
  • Highest accuracy subgroup: sensitivity 95%, specificity 97%, AUC 0.99.
  • Machine learning showed higher performance than deep learning; serum biomarkers outperformed plasma.
  • External validation improved specificity but reduced sensitivity.

Conclusions

  • AI algorithms demonstrate satisfactory performance for ovarian cancer diagnosis via blood biomarkers.
  • AI holds potential as a future diagnostic tool, possibly reducing unnecessary surgeries.
  • Future research should prioritize external validation and deep learning integration into AI diagnostic models.

Keywords: