Maria Bretones, biomedical engineer at TIC Salut Social from Catalonia, Spain who presented “Systemising the adoption of Artificial Intelligence in public healthcare systems: the Catalan approach”. Maria introduced the Catalan Health AI Programme which embraces the entire lifecycle of AI tools, from conceptualisation to implementation. The main programme objective is to implement AI tools that bring systemic value, i.e. improving the care provided to citizens and supporting healthcare professionals.
I will be H talking about how we are trying to systemize the adoption of artificial intelligence uh here is um the content of my my presentation I will start talking a bit about the context and then I will go directly to the main AIS of this AI strategy in in
Catalonia so the health AI program was approved on March of this year and and it was born with a mission of creating an enabling environment for Innovation that could improve care for Citizens and as well as supporting healthc Care Professionals we have the vision of leading the implementation of AI
Solutions that contribute uh to healthcare quality and the sustainability of our Health Care system and respecting of course the values of transparency efficiency Innovation Etc which are the benefits of creating this kind of of programs we believe that uh we are promoting an AI ecosystem in our field of Health we also are
Participating in the promotion of data quality with this program we also want to allow an strategic alignment in in AI policies we also want to guarantee equity in the access of these AI Solutions and also we want to contribute to the systems efficiency so now that we have learned
About the the context I’m I’m going directly to the main axes of of this program there are three main axes that are research and Innovation evaluation and Healthcare System implementation each acts has their own strategic actions and I’m going to go directly to to the first one the the first strategic
Action would be launching AI challenges the main aim of this strategic action is to cover the needs of the healthcare system that could be solved through the the application of AI nowadays we have four ongoing challenges uh the first one on diabetic retinopathy then we have another challenge on cheex x-ray ident well
Identifica the main findings in test x-ray then we have another two challenges in the scope of Dermatology and Drug prescription and medicines here you can see uh an like an schema of how we are imagining these challenges and as you can imagine adopting a an AI solution at a systemic
Level is complex and involves several institutions and and professionals and what we well what what the challenges try to do is um first first of all identify the needs and we go through a challenge proposal development and and approval and once the the executive committee of the program accept accepts this challenge we
Start involving professionals and experts in a specific working group of this challenge then we start with the procurement process and and we um we decide the the the specifications of this procurement process normally we start with a preliminary Market consultation to know a bit more about the context and and the possibilities
The market can offer then once we publish these specifications we start receiving the proposals and we evaluate them a differential Point here is that as we are talking about AI algorithms we need to validate those algorithms with our own data that’s why there’s a a validation step uh with our own um with
The Healthcare System data sets to validate those algorithms then um we adjudicate the contract and we would start the deployment the deployment stage once the solutions are deployed we would start this tracking uh step where we are continuously assessing these Solutions and and tracking its its Evolution and monitoring
Them another strategic action in these acts is supporting local AI initiatives the main aim of this action is promoting those initiatives in the Catalan ecosystem that could bring high value to to the whole system how we do it we Mentor projects during their definition and execution and we also uh
Publish good practice guidelines now we have two guidelines published one on good practices for code development in the scope of AI for health and another one h on explainability and soon we will publish two more H about C marking and data protection also with this strategic action we want to define the
Requirements for the implementation of AI we are uh aiming for uh implementing trans worthy AI tools and here uh we take into account the future a AI principles future AI it’s uh as you can see on the map an international framework uh that is being developed by um a multidisciplinary team where the
Word future each letter aims for one principle of trustworthy AI that has to be fair Universal traceable usable robust and explainable then the third strategic action of of these first acts would be the health Observatory here we work in two main directions in one hand we have the
Register of AI tools where we analyze the level imple of implementation of AI and the maturity of the solutions we identify and then we also want to act as a rather of the latest innovation in AI at an international level here I would like to show you some
Of the of the first indicators that we are collecting in in the observatory nowadays we have 145 AI tools registered by 79 different entities and we classify them um according to the technology Readiness level and we also classified this technology Readiness level in in four different phases research prototyping development and
Deployment the main specialty medical specialty is oncology but we found out that more than 20% of the tools are transversal that means that we could apply them in in different Medical Specialties we also wanted to highlight that it depends on on the maturity level but not all the tools consider
Interoperability and standards with in the with in the future could uh bring difficulties when integrating these Solutions and we also have 19 C Mark tools uh registered mainly class 2A then I would like also to highlight that only 42% of the tools that we have registered H use explainability techniques mainly variable variable
Importance but here on the table you can see other combinations we have encountered and also regarding input data most tools use images and some of them combine them with tabular data here also you you can see other combinations we have seen and then if you remember the program had three different axes the
Second one evaluation here we want to evaluate the solutions before their implementation and uh before the implementation at a systemic level and we are developing a methodology for that and also we are developing or designing a monitoring system uh for once these Solutions are already implemented this
Monitoring system will will allow us to build indicators reports and uh plan periodical retraining if necessary and then we have the implementation acts Where We Are working to integrate Solutions and here we would also include data governance and uh we are also working on training uh of healthcare professionals
We are developing an an AI plan depending on on the needs and and the profile and uh we are organizing different different talks and and content to to keep Healthcare professionals up up to date in in our in our area and the last strategic action here would be communication we
Are now designing a new website for for the program the visual identity and we are participa participating in congresses like like this one so that would be what I wanted to tell you thank you very much for attention