Interview: “I am very interested in working to ensure that new artificial intelligence solutions are adopted with the necessary safeguards and can have an impact on people’s health.”

Maria Bretones is 23 years old and since September has been part of the Artificial Intelligence team at the TIC Salut Social Foundation, where she also did a four-month internship. She has just presented her final degree project in Biomedical Engineering, based on developing an explainability method for multimodal artificial intelligence in the field of health. We spoke to her to find out more about her work and her first experience of working in the world of artificial intelligence.

Why did you study Biomedical Engineering?

When I had to choose my degree I was not really sure what I wanted to do. I decided on Biomedical Engineering at Universitat Politècnica de Catalunya because I wanted a technical degree and I was interested in the field of health. In addition, despite it apparently being quite a narrowly-focussed degree, I felt it covered many different areas while allowing for a lot of different specialisation options.

What did your final degree project consist of?

My final degree project consisted of finding and testing a methodology that would allow us to obtain explanations of how models based on more than one modality of data act. I specifically focussed on models based on medical images and tabular clinical data, since these two modalities cover most of the patient health data generated. I chose this topic because the foundation was already working on applying explainability methods but we had not explored this area in the context of multimodal artificial intelligence (AI). This added a degree of complexity and could be considerably useful.

What are your conclusions?

The conclusions of the project can be summarised in two main ideas. Firstly, it is expected that as the field of multimodal AI becomes consolidated, so will the associated explainability methods. In the current context, the method proposed in the project makes it possible to obtain explainability results with a simplified approach that can be useful in multiple use cases. Secondly, it has confirmed the importance of developing explainability methods for any AI solution to enable and expand its understanding, improvement and reliability and facilitate its subsequent adoption by health systems and professionals who may use it. So if we are moving towards developing multimodal AI models, it is essential also to develop compatible explainability methods.

What did your experience at the foundation add to your project?

Since I started at the foundation a year ago, I have been able to learn about the entire ecosystem surrounding AI and the aspects that must be taken into account to develop solutions applicable to health. It was here that I came across and learned about the concept of explainability and its relevance, especially in our field. So the first months of my internship at the foundation were decisive in choosing the project topic. After starting and narrowing down the project topic, the work I have been doing at the foundation has helped me to better understand the needs of the field, as well as the existing solutions. This has allowed me to know and explain the entire context and justification for the work better.

What are you currently doing at the foundation and how do you assess your experience there?

I am currently working in the Technical Office of the Health/AI Programme, led by the foundation’s AI Department. At the office we manage the documentation arising from the programme’s activities and tasks. We are also responsible for establishing the necessary contacts with the teams involved in their performance. In recent months, I have mainly worked on the Preliminary Market Consultation for the Diabetic Retinopathy Challenge, which we completed in February. This included individual sessions with various organisations in the sector. My experience has been very positive, as it has given me a very broad view of the projects being carried out, which are very varied and involve many different professionals. It is a highly enriching experience.

Which area would you like to specialise in or continue working on in the future?

When I did my final degree project I discovered the area of multimodal AI, which I believe can add a lot of value to the field of health. I would like to deepen my knowledge in this area. I am also very interested in working to ensure that these new artificial intelligence solutions are adopted with the necessary safeguards and can have an impact on people’s health.