Finalist of the Alsatian edition of “My thesis in 180 seconds”, Stanislas Demuth, doctoral student in bioinformatics (Inserm – University of Strasbourg) works on the theme of the application of computer technologies to the care of multiple sclerosis (MS ). We spoke with him about his project, about augmented intelligence and AI in health, precision medicine tools and the digital transition that the sector is experiencing.
1) Your thesis focuses on computational approaches for precision medicine in multiple sclerosis, can you tell us about your project?
My PhD aims to develop precision medicine tools in multiple sclerosis, an autoimmune brain disease that affects around 120,000 French people. It is a question of exploiting digital resources and technologies to assist medical decisions. My reflection therefore lies between technological development and exercise. I regularly alternate between the practice of medicine and scientific research because of a double course in medicine and science. My field of scientific interest has evolved during this journey. First focused on immunology, medical practice made me more aware of applied research and in particular the challenges of the digital transition of the profession.
” Computers have always appeared to me as a tool for carrying out tasks more efficiently or extending one’s field of competence, whether for associative, audiovisual or now medical projects. ».
With my thesis supervisors, we therefore designed my doctoral project as a co-direction and reconciliation of the clinical view (business) of Professor Jérôme de Sèze, expert in multiple sclerosis, and the bio-informatics view (technical) of the Professor Pierre-Antoine Gourraud, expert in precision medicine. A three-year availability for a science thesis is rare in medicine, where the course is already long, but I am comforted by the feeling that this subject corresponds to me.
2) How does AI fit into your work?
It is more about augmented intelligence than artificial intelligence. That is to say, it is a question of broadening the field of competence of the doctor rather than completely automating a precise cognitive task. The digitization of health data has exploded the volume of potentially exploitable data, the web makes it possible to share resources and the free software movement is spreading technology, particularly in the form of python modules. But moving from massive data to genuine “Big data” requires developing tools to exploit these elements in the action of the profession.
For this, the data architecture is more important than the performance of the predictive models. The simple possibility of accessing relevant data for a patient and making calculations on the fly during a consultation would already be a small revolution.
Imagine that I am in consultation with a patient who has recently declared multiple sclerosis. We have many treatments that slow down the progression of the disease, in this case, the progression of a motor, urinary, cognitive disability, etc. These treatments are more or less strong with a risk of adverse effects going hand in hand. Some patients need strong treatments because their disease will be aggressive, others may have disease that remains dormant for years before reactivating. Which are which?
“This is the challenge of precision medicine: identifying risk profiles to proactively manage the disease and identifying benign profiles so as not to cause unnecessarily adverse effects”.
Currently, we are already trying to adapt the treatments to the patient’s profile. We build on our previous experience and on the results of epidemiological studies that look for risk markers at the population level. Today, when a doctor announces to a patient that he has a 10% chance of being in a wheelchair within 5 years, he is in fact extrapolating the average risk in the population or a sub-population of a study. This is an approximation of the individual patient’s risk. It becomes more approximate when there are several relevant studies to cross-check and the doctor extrapolates mentally. As the evolution of multiple sclerosis is very variable from one patient to another, this level of precision is not sufficient. However, this is the best possible granularity as long as the experience is only shared at the population level.
The approach of the PRIMUS project, in which I participate, is to share the experience on an individual scale by reusing data from national scientific registers or therapeutic trials. Neurologists who are experts in the disease do indeed have a culture of collecting data from patients that they have been following for years. We are developing software that allows on-the-fly access to individual data relevant to the patient seen in consultation: in this case, data from patients similar to mine. Thus, the doctor of tomorrow will explain to the patient that out of the 10,000 patients in a reference database, 100 have had their profile in the past and that those treated with such a drug have had the best evolution. And he can do this by showing him the analysis made on the fly on his computer screen. We would therefore be in a situation where the machine reproduces the mental process of the doctor in a more granular, more efficient, more factual way, and explained while serving as a support for discussion between the doctor and the patient.
The machine therefore underlies a form of trio dialogue between the doctor, the patient and resources: similar patients (known as reference patients), predictive models, knowledge bases, etc. The algorithm identifying relevant reference patients can be as simple as matching. This simplicity promotes physician adoption. There is also a challenge of protecting the identity of people whose data serves as reference patients.
“We hope that an innovative anonymization technique will prevent even indirect re-identifications. In summary, we use the machine to delegate simple and laborious tasks: querying a database, making a calculation or producing a figure”.
At this stage of the work, the decision support is a simple description of the evolution data of these reference patients without involving a complex predictive model. In this, we are in an approach of augmented intelligence rather than artificial intelligence. Of course, the two are not mutually exclusive! Once this data architecture has been established, we will be able to address more complex tasks such as pattern recognition by machine learning models strictly speaking. It will also involve inferring prognoses in rare situations where the number of relevant patients is not large enough to be reliable. Matching reasoning may have only limited predictive performance and will need to be refined by predictive models that can take into account more complex, but less explainable, relationships. But we believe that by then, the culture of “hands-on” decision-making will have progressed.
3) How do you approach post-doctorate? What are your professional goals?
I will do a new work-study program, this time from research to the practice of medicine, for the final year of my medical course. This double course will have lasted 14 years. After that, I aspire to a mixed activity sharing time for exercise and research: a hospital-university career, the theme of which will be to contribute to the digital transition of neurology.
There are so many things to do that it will probably go beyond neurology, which could open up great prospects for both doctors and patients.
4) How do you see the evolution of health in the face of the rise of AI? How do you imagine the sector in 5-10 years?
I see several gaps. Much effort is currently being made in technological development, less in the development of tools for the action of the profession. On the one hand, some players talk about digital twins, neural networks to predict the evolution of rare cancers, etc. Often these technologies rely on research rather than routine data. On the other hand, in everyday life, doctors are still being overwhelmed by administrative burdens requiring multiple manual entries of the same data, emergency services are saturated with polypathological patients whose care is simply made time-consuming fact that their medical records are not shared efficiently, the ergonomics of hospital and office software still leaves something to be desired, hence the inflation of computer time at the expense of medical time.
This brings me to another opposition between, on the one hand, startups that develop digital tools preferentially for chronic diseases, whose care is subsidiary and relatively standardized. We must recognize the explosion of scientific knowledge that calls for precision medicine and hyperspecialization. On the other hand, the doctors who seem to me the most demanding and the most inclined to adopt these tools are those involved in primary care (general medicine and emergencies), whose exercise is cognitively demanding, because it is not affiliated by nature. They do not have the possibility of managing this complexity by hyper-specialization since the problems are intricate in the same patient. This preferential offer for affiliated care surely stems from a survival bias of start-ups in these markets.
The emergence of standardization of clinical data and their integration will be the trigger for digital medicine, like radiology, whose imaging format has been standardized for a long time: DICOM. Once we have defined the data we are working on, we can start doing machine learning may impact practice. I see the actual AI models as a modeling layer on top of this medical “big data” architecture.
“For me, the real challenge for the profession is to take the initiative in digital technology to design – or at least pilot – the development of the tools necessary and adapted to the field to manage contemporary medical daily life”.
Although, under current conditions and understandably, many doctors are resistant to the idea of tools requiring more time on the screen, a fortiori during the consultation, I think it is a question of ” toolbox “. Screen time will decrease with efficient tools. The appropriation of digital by our profession will open up prospects for data architecture allowing portability, the automation of time-consuming tasks, the synthesis of medical records, perhaps a renaissance of expert systems, then the addition of a layer of artificial intelligence strictly speaking: augmented intelligence, in short.
We would like to give thanks to the writer of this article for this outstanding material
AI in health, precision medicine and multiple sclerosis: interview with Stanislas Demuth, doctoral student in bioinformatics
Discover our social media profiles , as well as other pages that are related to them.https://yaroos.com/related-pages/