Incubator for AI validation

AIDA runs an incubator that assists in the startup of national validation platforms for AI in diagnostic imaging. This is a special initiative within AIDA addressing healthcare’s need to validate AI solutions, that is, to help care providers to take their responsibility to control that the solutions used are safe and effective.

Background

Innovators in healthcare face many difficult challenges when AI solutions are to be brought from the development labs to actual use at a given care provider. How to find out whether the solution is safe and effective under the local prerequisites? Where should the solutions be introduced in the existing workflow and what steps need to be adapted? To answer these questions, clinical and locally grounded validation is needed, beyond what the vendor has done to get the product certified. Such AI validation is a new area for healthcare and largely a capability that so far is lacking.

Incubator

Our incubator aims to assist the startup of national platforms for AI validation in diagnostic imaging. In the current phase, focus is on defining three pilot platforms in mature areas. The first one targets mammography screening and this will be developed to be ready for complete deployment at the end of the current phase, in the beginning of 2024. The objective is that the mammography platform from then onwards should work independently and sustainably within the regular healthcare structures.

VAI-B is introduced in further detail at Regional Cancer Centers in collaboration (in Swedish): Nationell valideringsplattform för AI inom mammografiscreening

The other two pilot platforms, one in radiology and one in pathology, are to be prepared for quick implementation after the current phase. Apart from the pilot platforms, the purpose of the incubator is to build a general capability in Swedish healthcare to efficiently create validation platforms for additional areas as needs arise. This is relevant both for other areas in diagnostic imaging as well as for medical technology in general.

Strategy

A cornerstone in our work is that healthcare needs to be in the driver’s seat already during the buildup of incubator and platforms, to facilitate speedy translation to regular operations after the incubator period. The steering group of the platforms is, therefore, composed of experts from the National Program area (NPO) Medical Diagnostics and Regional Cancer Centers in collaboration (RCC). Another cornerstone is strong research groups specializing in AI validation for mammography screening (Karolinska Institute and Lund University) that lead the first platform development. In contrast, the care providers to be pilot users of the platforms will be sought among smaller regions without in-house academic expertise, to ensure that the incubator contributes to equality of care across the country.