Assessing the Clinical Validity of Blind-Sweep Protocols for AI-Driven Ultrasound
Research Group:Status:Active (AI model training and protocol assessment in progress)Funding Agency:Partners:
GE HealthCare
Paropakar Maternity Hospital-Nepal
This project ensures the reliability and clinical usefulness of the blind-sweep ultrasound method. Even with standardized sweeps, real-world variations in technique or incomplete scans can occur. AI tools are being developed to automatically check each scan for quality and completeness, ensuring no critical information is missed.
Background
Task-shifted ultrasound can improve access to pregnancy care, but variations in scanning quality reduce reliability. Standardized sweep protocols alone may not ensure complete, accurate scans. Automated quality checks are needed to guarantee clinical usefulness.
Research Aim
To develop AI tools that evaluate each blind-sweep scan for clip-level quality and full-scan completeness, ensuring scans capture all critical areas and meet clinical standards.
Outcomes
The project has developed AI algorithms that detect issues like probe flipping, lost contact, reversed sequences, or noise. Reliable scans can later support clinical predictions such as gestational age, fetal position, and number of fetuses, making task-shifted ultrasound safer and more dependable.