Performance Metrics for Deep Learning Models in Aortic Dissection Imaging

A systematic review and meta-analysis aggregating results from 48 studies on deep learning models for aortic dissection segmentation and diagnosis, published up to November 3, 2024. The data, compiled by author Yichen Zhao, includes performance metrics such as Dice coefficients and pooled sensitivity/specificity for models based on CT, CTA, and ECG imaging. The dataset is a 22.1 KB document summarizing the findings of the review.

Use Cases

• Benchmarking segmentation model performance based on reported Dice coefficients for false lumen, true lumen, and entire aorta.
• Comparing diagnostic accuracy of CT-based, CTA-based, and ECG-based deep learning models using pooled sensitivity and specificity metrics.
• Evaluating the potential of AI as a clinical assistive tool by comparing model performance (0.79 sensitivity, 0.95 specificity) against clinician performance mentioned in the review.

Strengths

• Aggregates results from 48 identified studies, providing a consolidated evidence base.
• Includes specific performance metrics: mean Dice coefficients (e.g., 89.2% ± 4.4% for false lumen) and pooled sensitivity/specificity with 95% confidence intervals.
• Risk of bias was evaluated using the standardized QUADAS-2 tool.

Limitations

• The dataset is a 22.1 KB summary document (DOCX); the underlying raw data or column-level details are not provided.
• Row count and individual study data points are unknown, limiting direct analysis.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Systematic review search across Cochrane Library, PubMed, Embase, and Web of Science.

Collection Method

Meta-analysis of published studies identified through a comprehensive literature search.

Time Range

Studies included up to November 3, 2024.

Freshness

Last updated 2026-04-14 04:21:48; freshness should be verified.