CSAIL, the MIT Computer Science and Artificial Intelligence Laboratory, the MIT Institute for Medical Science and Engineering, the National Academy of Medicine and the Health Data Club organized last year’s symposium “Artificial Intelligence and Medicine: promises and limits”. Given the success of this event, they decided to repeat the experience with the support of the French Embassy in the United States. The second edition was held on the afternoon of October 20th in the premises of the Académie de Médecine in Paris.

The first edition of the symposium, held in May 2021, also focused on the promises and limits of AI applied to medicine, and brought together nearly 600 participants, including a large number of researchers from Europe and the United States.

This year’s themes focused on the challenges of access to data and the concrete applications of AI in medicine, whether in terms of clinical trials, brain-machine interfaces or the discovery of new drugs.

The afternoon was organized into three conferences that brought together 420 participants. Patrice Tran Ba Huy, President of the French National Academy of Medicine, gave a welcome speech before Bernard Nordlinger, President of the Commission on Digital Medicine, French National Academy of Medicine, and Daniela Rus, Director of the CSAIL and Vice-Dean of Research at the Schwarzman College of Computing, MIT, opened the symposium.

Bernard Nordlinger highlighted the stakes of the Symposium:

“by bringing together the players at the forefront of AI in health, 9 rue Georges Pitard in Paris, this half-day event demonstrates the importance of forging international collaborations on these topics, in order to amplify research opportunities through health data.”

Daniela Rus, for her part, recalled the potential of AI and the importance of data in the field of health:

“AI is an unprecedented vector for health and is already revolutionizing its uses, improving decision-making thanks to data. The challenge now is to coordinate ourselves to further improve patient health.”

Interoperable and open source data

The first session focused on data access at the national or international level. The lack of uniformity in data models and the different laws concerning their protection in different countries make the use of databases more complex.

The speakers in the first session highlighted the need for a common governance of data. Leo Anthony Celi, Senior Research Scientist at MIT’s Institute for Medical Engineering and Sciences (IMES), stressed the importance of agreements on data interoperability (one of the moderators was Elazer R. Edelman, Director of IMES and Edward J. Poitras Professor of Medical Engineering and Sciences at MIT).

Using quality data and avoiding bias when building algorithms

The speakers in the second session emphasized the fact that each of the actors (data producer, health care personnel or regulatory authority) has a role to play in ensuring the quality of the data to avoid any bias in the construction of the algorithm and the importance of working together for the benefit of the patient.

Farhad Rikhtegar Nezami, a researcher at Brigham and Women’s Hospital at Harvard Medical School, said:

“We need to keep in mind that the use of AI in healthcare is a journey, not a destination. Ultimately, it is for the benefit of the patient that we use these tools. The important thing for us now is to prove the benefit of AI and show that we can refine the process.”

Concrete applications of AI to medicine

The third session discussed the concrete applications of AI in medicine, both in terms of discovering new drugs, improving clinical trials and the possibilities for the brain-machine interface.

Nicolas Do Huu, Co-founder and Chief AI Officer at Iktos AI, a French start-up created in 2016 specialized in the development of AI solutions for research in chemistry, and in particular in medicinal chemistry and new drug discovery, spoke during this third part of the symposium. Iktos develops a proprietary and innovative technology using generative deep learning models, which allows, from existing data, to design optimized molecules in silico on all objectives of a molecule discovery project.

He mentioned different AI technologies for medicine: generative modeling, predictive modeling of the activity of the studied molecule, deep reinforcement learning, virtual screening and 3D/4D simulation. For him, AI brings innovation, he thus stated:

“For the creation of drugs, it could be a revolution analogous to the one the automotive industry is facing today.”

While large-scale molecular analysis will facilitate new drug discovery, the use of real-life data will improve the efficiency of clinical trials, and the brain-machine interface will see its applications increased tenfold by AI.

The speeches of American and European experts in AI in health have confirmed the growth of the challenges of health data exploitation in Europe and internationally.

Translated from Retour sur la seconde édition du symposium « Artificial Intelligence and Medicine : promises and limits »