Dados do Trabalho

Título

Diagnostic Accuracy of Artificial Intelligence for the Evaluation of Pericardial Effusion with Ultrasonography: A Meta-Analysis

Objetivo

Ultrasonography (USG) is highly operator-dependent and the role of Artificial Intelligence (AI) in enhancing diagnostic accuracy for pericardial effusion (PE) is emerging. This meta-analysis aims to evaluate AI's accuracy in detecting PE from USG images.

Métodos

We systematically searched PubMed, Embase, and the Cochrane Library for observational or randomized studies assessing AI's diagnostic accuracy for PE via USG. Data extraction was performed following Cochrane guidelines. Pooled accuracy and area under the curve (AUC) with 95% confidence intervals (CI) were calculated using a random-effects model. Sensitivity analysis was performed to assess for bias.

Resultados

A total of three studies met the inclusion criteria. Potter (2023) utilized Point-of-Care Ultrasound (POCUS), while Cheng (2023) and Nayak (2020) used standard echocardiography. Total sample included 1934 subjects, among images and patients. All studies included images using the four-chamber apical window (FC), with Nayak also analyzing 800 subcostal window (SC) images. Summary accuracy was 0.89 (CI 0.88–0.90) and AUC was 0.92 (CI 0.87–0.97). Sensitivity analysis was performed excluding the SC dataset resulting in a summary accuracy of 0.90 (CI 0.88–0.92) and AUC of 0.91 (CI 0.83–0.98).

Conclusão

AI models demonstrate high accuracy for detecting PE from USG images. There was significant bias identified in the sensitivity analysis performed. The observed heterogeneity may be attributed to threshold effects and varying imaging techniques. Further research exposing the models to a larger sample size is warranted to validate the findings.