Artificial Intelligence (AI) is increasingly becoming a standard of care in healthcare, yet concerns about its regulation and governing principles persist. Clinical AI started as a solution to individual hospital challenges, but its range has since expanded, making a profound impact across the healthcare sector. Despite various regulatory bodies and organizations providing guidance, a singular regulatory body for AI is currently non-existing, leaving the healthcare industry in need of enhanced oversight.
To provide the necessary guidance, “Evaluating Commercial AI Solutions in Radiology” or ECLAIR was introduced. It recognizes the critical role of radiology in adopting new technologies, suggesting ten pertinent questions for radiology departments to consider while choosing AI vendors. This list discusses key aspects, such as defining the clinical need, evaluating benefits and risks, and understanding the validation and performance of AI applications.
Firstly, departments should define the specific problem they want AI to address, identify the target users, and establish measures of success post-implementation. Next, they need to balance potential benefits against possible risks considering how the AI solution might impact clinical outcomes and workflow efficiencies.
Third, understand the AI solution’s validation process, ensuring it is rigorous, independent, and thoroughly assessed for its performance across relevant patient populations and imaging modalities. The fourth question revolves around the solution’s interoperability and integration within the existing workflow and system infrastructures.
Fifth, it’s crucial to consider IT requirements, engage the IT team early on to mitigate any potential challenges. The sixth question involves ensuring regulatory compliance with medical device and data protection laws in the target country – for instance, FDA clearance in the US and CE certification in the EU.
The seventh involves conducting a Return on Investment (ROI) analysis to determine economic viability. Next, consider the future maintenance and support to ensure continuity even as the IT department, as well as system requirements, evolve.
The ninth factor is evaluating the availability of user training and support mechanisms to facilitate effective adoption of the AI solution. Lastly, there is the need to develop protocols for managing potential errors and ensuring constant improvement through post-implementation surveillance.
ECLAIR guidelines offer a fundamental understanding of what practitioners need to look out for while considering AI solutions, thus providing much-needed direction in the rapidly evolving field of AI in healthcare.
