The future challenges of artificial intelligence in healthcare

Digital in the hospital is making great strides. Some algorithms are already better able to diagnose lung cancer or analyze a mammogram than a doctor. This technology promises immense progress in the future, provided certain challenges are met. Various protagonists and analysts of the sector gathered during the round table”Digital, artificial intelligence, doctors and patients” of the Global Health conference in Strasbourg, analyze the future challenges that the world of technology and the world of health will have to meet hand in hand.

Data yes, but quality

To have good AI, it has to work with a good algorithm. And a good algorithm is built with good data“, explains Bernard Nordlinger, a member of the National Academy of Medicine and a specialist in artificial intelligence in healthcare. Health data, collected from hospitals and healthcare facilities, must be calibrated, relevant and “cleaned” of any parasitic detail that would disturb it the interpretation.”We recall the artificial intelligence Watson, created by IBM, which was very disappointing, in particular because the health data used was not ideal.”

The scientists who worked on this program explained that they had great difficulty understanding patient records: acronyms to detail, errors to correct, abbreviated sentences. Each piece of information must first be formatted to be correctly analyzed by the system. “The problem was the same during the Covid-19 crisis. Artificial intelligence could not find its place because the data generated at the time was not of high quality“, regrets the specialist. On the other hand, when the data is accurate and well formulated, they can also have their place in a clinical trial. Today there are control groups, i.e. the arm of a clinical trial that do not take the treatment, composed only of synthetic data made with artificial intelligence.A promising lead.

Cooperate internationally

After their development and before their use in healthcare settings, algorithms undergo rigorous evaluations in clinical trials, such as during the development of a drug. “Validating these algorithms requires international cooperation“says Irene Buvat, director of the Translational Imaging in Oncology Laboratory at the CNRS and director of research at the Institut Curie.”We are carrying out many projects with European partners to build databases on which the algorithms will be developed. But the validation phase with other countries is also crucialThe specialist explains that the different protocols and criteria in place in each country present an opportunity to refine artificial intelligence, so that it can be implemented in hospitals internationally.

Some algorithms are built using health data collected in hospitals, at the patient’s bedside. “It is therefore necessary to be able to collect data in the hospital but also sometimes obtained from remotely connected objects“, explains Jean Sibilia, PU-PH in rheumatology at the University Hospitals of Strasbourg and the Faculty of Medicine of Strasbourg. “Physicians also need to be updated to ensure ethics for patients and reassure them. And finally, they must be able to analyze this data.” Objectives that can be included during initial training as well as later, in the context of continuing training. “Ethics and data protection are issues to be addressed both with senior doctors and with the younger generations who have grown up in the digital world.“, continues Jean Sibilia.

Cybersecurity must be there

Artificial intelligence means generating health data within health centers. But these are now in danger. Ransomware targeting hospitals has multiplied in recent years. These cyberattacks paralyze healthcare facilities and threaten the privacy of patients, whose “data” can be resold on the dark net.

We don’t ask if it will happen, but when”, confides Antoine Geissbühler, head of the cyberhealth and telemedicine service of the university hospitals of Geneva.When faced with vulnerable systems, we need to reduce the attack surface available to cybercriminals. We have ethical hackers and intelligent monitoring systems on our side, capable of detecting suspicious behavior before a massive attack. A bit like banks able to cross-check when the use of the credit card is anomalousIT updates to stay up-to-date and networking between hospitals are also part of the indispensable tools for knowing how to limit the risk of a cyber attack.

Leave a Reply

Your email address will not be published. Required fields are marked *