AI Ultrasound Video Dataset Released
A new open-access video dataset for maternal-fetal ultrasound, called Ultrasouno, has been released to help develop AI models. The goal is to create automated tools for intrapartum biometry that can support real-time clinical decision-making during labor. This initiative aims to improve fetal monitoring and risk assessment through data-driven technology.
- The Ultrasouno dataset was developed by an interdisciplinary research team at the University of Oxford's Department of Engineering Science and the Nuffield Department of Women's & Reproductive Health. The project, known as PULSE (Perception Ultrasound by Learning Sonographic Experience), aims to use AI to build computational models that mimic the visual and decision-making processes of expert sonographers. - A primary challenge this technology addresses is the subjectivity and inaccuracy of traditional intrapartum monitoring. Studies have shown that digital vaginal exams to determine fetal head position can be incorrect in 20% to 70% of cases, highlighting the need for more objective assessment tools. - AI models trained on datasets like Ultrasouno can automate the measurement of key indicators of labor progression, such as the angle of progression and fetal head station, with greater accuracy and reproducibility than manual methods. This has the potential to reduce errors related to operator experience and challenging clinical situations like significant fetal head molding. - For nurse-midwives, AI-powered tools can serve as a decision support system, freeing up time from routine tasks to focus on direct patient care, education, and emotional support. By providing data-driven insights, these technologies can help in the early identification of potential complications, allowing for proactive interventions. - Professional midwifery organizations, like the Australian College of Midwives, recognize the potential of AI to improve safety and efficiency in maternity care but also emphasize the need for midwives to be integral in the design and implementation of these tools. The Council of Deans of Nursing and Midwifery in Australia and New Zealand also supports the integration of AI into education to prepare future graduates for technology-enabled practice. - The adoption of such technologies is particularly relevant in the context of healthcare workforce shortages. AI can help optimize staffing and reduce burnout by automating repetitive tasks and streamlining workflows, which is critical as over 138,000 nurses have left the workforce since 2022 due to issues like stress and burnout. - In Virginia, recent legislation has expanded the scope of practice for Certified Nurse-Midwives (CNMs), allowing for greater autonomy. As of 2024, a CNM who has completed 1,000 hours of practice may practice without a written or electronic practice agreement, consulting and referring to other healthcare providers as needed. - The Virginia Board of Nursing requires that CNMs practice in accordance with the standards set by the American College of Nurse-Midwives (ACNM). The ACNM's standards and core competencies provide the framework within which new technologies like AI-assisted ultrasound would be incorporated into midwifery practice.
