Hormonal and Laboratory Predictors of Patent Foramen Ovale in 300 Female Patients

A retrospective study of 300 female patients (18–65 years) with cryptogenic ischemic events, assessing 43 clinical and laboratory variables to predict patent foramen ovale (PFO). The dataset was used to develop a logistic regression model, achieving an AUC of 0.990, and includes results from SHAP analysis and subgroup analysis by reproductive stage. The data was authored by Yao Zheng and shared on figshare in April 2026.

Use Cases

• Training machine learning models for PFO detection based on clinical variables like age, estradiol, and D-dimer.
• Investigating hormonal biomarker relationships with stroke risk across different reproductive stages.
• Validating feature importance explanations using SHAP analysis on the identified five key predictors.
• Conducting comparative analysis of model performance (e.g., logistic regression) using metrics like AUC, accuracy, and sensitivity.

Strengths

• Includes data from 300 female patients with a clearly defined outcome (60 PFO-positive cases).
• Model performance is quantified with specific metrics, including an AUC of 0.990 and accuracy of 95.6%.
• Analysis includes a prespecified subgroup analysis by reproductive stage (reproductive-age vs. menopausal).
• Feature selection identified five key predictors: age, estradiol, FSH, D-dimer, and LDL cholesterol.

Limitations

• Row count and column-level documentation are unknown, which may limit suitability assessment.
• The dataset is small (214.3 KB), indicating limited scope and sample size.
• Data may reflect bias inherent to the single retrospective study design and specific patient cohort.

Provenance

Source

figshare, author Yao Zheng.

Collection Method

Retrospective study of patients who underwent contrast-enhanced transcranial Doppler and echocardiography.

Freshness

Last updated 2026-04-17 04:12:15; freshness should be verified.