Can you briefly introduce yourself and Aidoc’s mission?
I’m Demetri Giannikopoulos, Chief Transformation Officer at Aidoc. Our mission is to harness clinical AI to empower healthcare teams, optimizing patient treatment while driving both economic value and improved clinical outcomes. By identifying and aggregating medical data, we enable care teams to operationalize the unexpected and work seamlessly always keeping the patient at the center.
My journey in healthcare began in 2003, coincidentally the same week HIPAA went into effect. My first job at Dictaphone was to read and interpret HIPAA regulations to understand their impact on the business. At the time, I had no idea this would set the foundation for a career deeply rooted in healthcare innovation and transformation.
In 2019, Aidoc became one of the first companies to receive FDA clearance for radiology AI. Given my background in this space, I was drawn to the company’s vision and joined the team. Since then, I’ve been focused on driving innovation and transformation across radiology, neurosciences, cardiology, vascular surgery and critical care.
How did you become involved with the CHAI model card initiative?
In 2023, we started receiving questions from our hospital and health system partners about the HTI-1 proposed rule intended to implement sections of the Cures Act around interoperability and clinical decision support, which initially suggested that ONC-certified health IT vendors (primarily EHRs) be responsible for transparency requirements of AI use in healthcare.
We were asking ourselves, how can we show health systems that we comply with this draft law for AI and are in accordance with all the technical qualifications, despite not being directly subject to HTI-1? How do we make AI governance easier for not only the AI solution provider, but the health systems receiving and evaluating these technical forms that vary in complexity and lack standardization.
This led me to connect with CHAI and participate in the workgroups, specifically the model card workgroup, where we developed the initial draft versions of CHAI’s applied model card, which has since been iterated on and used as a primary tool in multiple Aidoc procurement decisions.
What value does the CHAI model card bring to providers of AI solutions in healthcare, and how is this played out in the real-world?
The model card offers a streamlined way to showcase the capabilities, limitations and compliance aspects of our AI solutions similar to the nutrition labels you see in everyday life. It helps build trust with healthcare providers by providing transparency beyond the typical “procurement through PowerPoint” approach. It allows solutions providers to clearly demonstrate how they comply with regulations like HTI-1, even if they are not directly subject to them.
Part of the value CHAI provides is the ability to move fast – 90 days after beginning the model card working group, we finalized the first completed draft from Aidoc and presented our model card in a live clinical environment to UMass Memorial Health’s AI Governance Committee. When we initially received a 30-page long version of the deployed CHAI Responsible AI Checklist to fill out, Dr. Elisabeth Garwood from UMass Memorial Health also shared that a model card could serve as a welcomed option in their governance process. Using the model card led to a very productive and comprehensive conversation for all who were involved.
How do you see the model card fitting into the broader landscape of AI governance and trust-building in healthcare?
When HTI-1 was finalized, we started receiving AI governance questionnaires, ranging from three to 75 pages of questions from health systems that we needed to complete– on one occastion, we received a few different governance forms totaling 125 pages, before completion. These forms not only take a lot of work and time for the solution provider to fill out, but, given the scale, the review process can many times take months for the senders to review.
At Aidoc, we use the model card to accelerate adoption of our solutions and gain trust with AI governance committees through the sharing of streamlined, rich, technically accurate and relevant information. It helps everyone get down to the real core issues of a discussion instead of working through tons of superfluous steps.
I believe the applied model card is a crucial step toward enabling innovation in healthcare AI while simultaneously building trust and providing a level of clarity throughout procurement and compliance. It’s a tangible tool that helps bridge the gap between regulatory requirements, innovative AI offerings, and the needs of healthcare providers—all while allowing for easier implementation at scale for successful tools.
When deploying Aidoc’s completed CHAI Applied Model Card, what is the feedback?
The feedback has been overwhelmingly positive—unsurprisingly, most people really enjoy experiencing new efficiencies and engaging in better processes. We’ve frequently heard that it’s clear and to the point, packed with the right details. It covers the intended use, key specs, and near everything a health system needs to understand how the AI fits into their ecosystem.
What future developments are you looking forward to?
I’m excited to collaborate with CHAI on an executive brief that delivers high-impact insights in a clear, digestible format—giving healthcare leaders the key takeaways they need without getting lost in technical complexity.
Another major initiative my team is driving is the completion of applied model cards for each of our 18 FDA-cleared algorithms. These model cards are a crucial tool for building trust with governance committees and accelerating adoption. When a solution is effective, model cards help scale its impact by providing transparency and reinforcing confidence in its performance. I look forward to leveraging this approach at Aidoc to further drive trust and innovation across healthcare.
Media Inquiries: chai@12080group.com