Meta-Analysis of Deep Learning Models for Aortic Dissection Segmentation and Diagnosis

A systematic review and meta-analysis of 48 studies published up to November 3, 2024, evaluating image-based deep learning models for aortic dissection. The document, authored by Yichen Zhao and last updated in April 2026, reports performance metrics including Dice coefficients for segmentation and pooled sensitivity/specificity for diagnosis across CT, CTA, and ECG modalities.

Use Cases

• Benchmarking new segmentation models against reported mean Dice coefficients for false lumen, true lumen, and entire aorta.
• Evaluating diagnostic AI performance against pooled sensitivity and specificity metrics for CT, CTA, and ECG-based models.
• Informing clinical tool development by comparing model performance to clinician performance metrics cited in the review.

Strengths

• Includes quantitative results from 48 studies, providing a substantial evidence base.
• Reports specific performance metrics: mean Dice coefficients (e.g., 89.2% for false lumen) and pooled sensitivity/specificity with confidence intervals.
• Compares model performance to clinicians, with pooled specificity of 0.95.

Limitations

• The dataset is a 2.5 MB DOCX file containing a review article, not the underlying raw data from the 48 studies.
• Row count and column-level documentation for any underlying data are unknown.
• Data may reflect publication bias inherent to the systematic review's source studies.

Provenance

Source

figshare, authored by Yichen Zhao.

Collection Method

Systematic review and meta-analysis of studies from Cochrane Library, PubMed, Embase, and Web of Science.

Time Range

Studies identified up to November 3, 2024.

Freshness

Last updated 2026-04-14 04:21:46; includes studies up to November 2024.