Objective:

To evaluate uFlowAM for fetus-level detection and visualization of abnormal intracardiac microflow patterns on early-pregnancy FCMI.

Approach:

• Study Design: Multicenter diagnostic accuracy study analyzing 650 early-pregnancy FCMI examinations from fetuses referred for suspected CHD or CHD risk assessment.
• Methodology: Standard clips were processed using microflow extraction, cardiac-cycle alignment, and self-supervised training to learn control fetal microflow patterns.
• Validation: Internal validation was performed to calibrate the operating threshold, followed by external testing and clinical utility assessment in a multi-reader multi-case study.

Key Findings:

• uFlowAM achieved an AUC of 0.94 in the internal cohort and 0.92 in the external cohort.
• Sensitivity was 0.92 and specificity was 0.88 in the internal cohort; sensitivity was 0.90 and specificity was 0.86 in the external cohort.
• Median reader-level AUC increased from 0.85 to 0.92 with uFlowAM assistance.
• Weighted kappa for subtype agreement increased from 0.62 to 0.78.
• Median reading time decreased from 78 seconds to 59 seconds.

Interpretation:

uFlowAM detected abnormal early-pregnancy fetal cardiac microflow patterns and improved reader consistency and efficiency.

Limitations:

• The framework should not replace a complete fetal echocardiographic examination.
• Examinations were excluded for severe fetal or maternal motion artifact and other quality issues.

Conclusion:

uFlowAM is proposed as an assistive second-reader tool for early fetal CHD assessment.