In this interview, we speak with Davidi Vortman, CEO of UltraSight, about their innovative AI-driven ultrasound technology and how it is transforming cardiac workflows. At the core of UltraSight’s approach is the commitment to empowering healthcare professionals across various settings to perform cardiac ultrasound effectively. With AI Real Time Guidance, clinicians with no prior ultrasound experience can capture high-quality cardiac images with ease and accuracy. This conversation explores UltraSight’s strategies for integrating AI into cardiac care, from enabling rapid adoption and scaling across healthcare environments to addressing regulatory and clinical challenges.
Moe Alsumidaie: How do you plan to scale training and adoption of your AI ultrasound tech in diverse healthcare settings, including low-resource environments?
Davidi Vortman: We designed our solution with end users in mind, recognizing the scarcity of specialists. UltraSight AI Real Time Guidance allows health care professionals with no prior ultrasound experience to become proficient within four to six hours of training. Clinical studies have validated this, showing that residents and nurses with no sonography background achieved diagnostic quality cardiac images with UltraSight AI Real Time Guidance. This rapid uptake is crucial for adoption of the technology in low-resource environments with limited access to specialists. We continue to enhance our training tools, offering online resources and practical on-site training to ensure that users can quickly and effectively utilize our solution.
Moe Alsumidaie: What challenges did you face in designing clinical trials to evaluate your AI system’s efficacy in real-world settings?
Davidi Vortman: Designing clinical trials for our AI system involved several challenges, particularly in ensuring diversity and real-world applicability. We had to demonstrate diversity in patients, users, and disease states. Therefore, we collected clinical trial data from 240 patients across three sites within six months, a significant logistical challenge. We ensured a diverse data pool and included subjects with various diseases, not just healthy individuals, to reflect real-world conditions. This approach required a transparent methodology for data collection and usability studies, demonstrating to FDA our AI’s functionality. The FDA now demands a clear understanding of the AI’s “engine,” including how algorithms are trained and validated, adding complexity to trial design. We ensured our trials accurately reflected the AI system’s performance in diverse clinical settings by addressing these issues.
Moe Alsumidaie: How did you ensure your AI system performs consistently across patient populations and disease severities?
Davidi Vortman: Ensuring consistent performance across diverse patient populations required careful site selection and data collection. We focused on cardiology centers with a high volume of diverse patients, such as the University of Chicago and Aurora St. Luke’s Medical Center. These centers perform thousands of echocardiograms monthly, allowing for study of a random and diverse patient population. We ensured the patients’ data was broad, including various diseases, genders, and BMIs. The study design, which did not include a follow-up requirement, simplified patient recruitment, reducing patient burden and allowing us to test the AI system’s performance across various demographics and disease severities. This approach aimed to ensure that our AI system could consistently deliver accurate results, regardless of patient differences.
Moe Alsumidaie: What measures did you implement to address the learning curve for novice users in your clinical trials?
Davidi Vortman: Addressing the learning curve for novice users was a key focus in our clinical trials. We ensured users had no prior ultrasound experience, and we provided consistent training. After training, users conducted eight exams independently, followed by expert debriefs. The learning was fast, with users gaining confidence after just eight scans, unlike traditional methods. UltraSight AI Real Time Guidance accelerates the learning process, allowing users to reach a high level of proficiency quickly.
Moe Alsumidaie: What regulatory considerations and adaptations have you made for international trials of your AI-guided devices?
Davidi Vortman: Expanding our trials internationally required careful consideration of regulatory differences in the U.S., Europe, and Israel and the varying approaches to AI regulation globally (for example, differences in scope and application of regulatory frameworks, and in enforcement mechanisms). We understand that these differences require us to adapt our trial designs and regulatory strategies accordingly. By addressing these regulatory nuances, we can ensure compliance and successful trial execution across different regions.
Moe Alsumidaie: How does your trial design account for potential biases introduced by AI algorithms?
Davidi Vortman: Mitigating potential biases in AI algorithms was critical to our trial design. The validation study included a comparison of a group of non-experts who used UltraSight AI Real Time Guidance, and a group of experts who did not use it. Each image was reviewed by five cardiologists using an objective rating scale, ensuring consistency and objectivity in image quality assessment. The cardiologists were blinded to the image operator, whether it was a non-expert, who was using our AI technology, or an expert, who did not use the AI guidance. The evaluating cardiologists were also blinded to each other’s scores and were not exposed to the others’ evaluation. UltraSight AI Real Time Guidance guides users to save only high-quality images, reducing the risk of bias. The AI provides real-time feedback, ensuring users capture the best possible images. This approach minimizes potential bias and ensures the AI system performs consistently across different patient demographics and conditions.
Moe Alsumidaie: What are your thoughts on AI and ultrasound technology’s potential in decentralized clinical trials?
Davidi Vortman: UltraSight’s cardiology workflow solutions hold significant potential for decentralized clinical trials, offering opportunities to reduce burden on patients who participate in trials and improve access to research. By enabling nurses without prior cardiac ultrasound experience to perform diagnostic cardiac ultrasound in clinical trials, we can reach more subjects at a lower cost, making research more accessible to diverse populations. This approach may also play a pivotal role for new drugs that require frequent cardiac ultrasound monitoring. The ability to broaden patient access by expanding the cardiac ultrasound user pool, could also potentially reduce the burden on healthcare systems, ultimately transforming the landscape of clinical trials and cardiac care.
Moe Alsumidaie is Chief Editor of The Clinical Trial Vanguard. Moe holds decades of experience in the clinical trials industry. Moe also serves as Head of Research at CliniBiz and Chief Data Scientist at Annex Clinical Corporation.
