AI in healthcare

Using artificial intelligence data implies the storage, manipulation, and sharing of large quantities of health data. However, administrative bodies categorize health data are sensitive, as they relate to the racial origins, political and religious opinions, state of health, sexual orientation, and other personal details of patients. Their use is subject to strict security regulations, and it is thus necessary to establish adapted security measures, such as strict policies pertaining to data access, dedicated storage spaces, and regular audits to ensure adherence to security measures. It is equally important to ensure the training and raise awareness of people who have access to these data, especially with regards to the security of the data and patient confidentiality.

How can patient confidentiality and health data security be guaranteed? What are the regulatory obligations and how can one adhere to them?

The development of AI-based medical solutions is based on the analysis and exploitation of health data for extracting the pertinent information. If these data are poor quality (incomplete, incorrect, contradictory, or obsolete), the results produced by the AI could be incomplete or wrong, which could have serious consequences in the context of care usage. It is equally important to take into account potential data biases in order to avoid their reproduction.

Data quality is a critical issue for implementation of healthcare AI, and it is thus essential to guarantee that the data used are reliable, precise and complete, in order to produce correct and useful results for healthcare professionals and patients.

How can we guarantee data quality in order to maximize the reliability and performance of medical AI diagnosis technologies? What processes should be implemented and which stakeholders should be involved?

While AI can bring solid support to healthcare professionals, its usage also poses some risks. It is key to train healthcare professionals in the appropriate usage of AI, whether it is in technical aspects (basis of AI, machine learning, and deep learning) or ethical aspects (algorithmic biases, data protection), but also in the manner of result interpretation. Training healthcare professionals in this manner is crucial for guaranteeing a responsible, ethical, and effective usage of healthcare AI.

How can we train and adapt healthcare professionals to guarantee the usage of relevant medical AI technologies? How can continuous training systems be implemented that align with the rapid market changes and the regulatory contexts linked to AI?

Development of AI in healthcare will not be possible without collaboration between different kinds of players. From one side, this means traditional healthcare players, such as institutional players (hospitals, regional health agencies), and pharmaceutical laboratories, which have medical competencies and clinical skills, as well as a better understanding of care challenges and needs. On the other side, digital technology companies (startups, GAFAM, telecom companies) can bring to the table their expertise in cutting-edge technologies, software development, and data usage. The collaboration between these two types of players is thus necessary for a more relevant and effective development of AI solutions in healthcare.

How can we promote the development of collaborations between players in medical AI? Which partners should work together based on the specifics of different contexts and projects?