Artificial intelligence in support of precision medicine
Jan 17, 2025 – By analyzing human-generated data, artificial intelligence can help refine medical diagnoses and adapt therapeutic treatments to individual patients. Nesma Houmani, Associate Professor at Télécom SudParis’s SAMOVAR lab, is working to meet the ‘concrete needs’ of medical teams in the field of neurodegenerative and chronic diseases. Focus on the researcher’s work on the occasion of the AI, Science and Society summit organized by the Institut Polytechnique de Paris on February 6 and 7, 2025.
At the Samovar laboratory, Nesma Houmani trains algorithms to identify pathologies or refine diagnostics from patient data.
It only takes an algorithm to bring together engineering and medicine, very much like the ones that Nesma Houmani, an Associate Professor at Télécom SudParis, designs to identify pathologies or refine a diagnosis. “The aim is to provide carers with a fine-tuned, accurate and rapid decision-making tool based on human clinical data. In other words, we want to help develop precision medicine,” says the researcher from the SAMOVAR* Lab (Distributed Services, Architecture, Modeling, Validation and Network Management).
The principle is simple, the implementation not so much. Firstly, the algorithmic models need to be trained to distinguish between several shades of the same pathology or the range of possible reactions to a treatment. The algorithms are trained, for example, to describe an image (MRI, X-ray, etc) and to interpret signals such as those from an electroencephalogram (EEG) or other patient-related data such as height, weight, age and date of admission to hospital. During the test phase, the algorithms are assessed in order to determine their performance and generalizability. “The outcomes of the algorithm’s analysis depend on the quality of the data, which may be degraded or imperfect”, underlines Nesma Houmani. Information may be missing from a medical file due to lack of time for an exhaustive examination, an electrocardiogram may contain background noise due to the machine, a patient may have moved during the recording of an EEG, etc.
Algorithmic intelligence
Therein lies the power of artificial intelligence, or ‘algorithmic intelligence’ as the researcher prefers to call it, which can process large quantities of data thanks to technological advances and deal with the great variability that is specific to the healthcare sector. “We now have a pretty extensive experience in the acquisition of human medical data, enabling us to improve the quality of the data we acquire,” says the Associate Professor.
For 12 years now, Nesma Houmani has been working in the healthcare sector with several medical teams from the AP-HP (Assistance Publique – Hôpitaux de Paris), including Paul Brousse, La Pitié-Salpêtrière and Charles-Foix hospitals, as well as the Centre Hospitalier Sud Francilien, the Genopole, and the UGECAM (Union pour la gestion des établissements des caisses de l’Assurance Maladie) rehabilitation centre in Ile-de-France, and the teaching hospital (centre hospitalier universitaire) in Lille. Her latest research focuses on the study of brain activity using electroencephalography, and gait walking analysis in the case of neurological diseases.
Researchers at Télécom SudParis are testing different methods, adjusting them and defining the relevance of the results obtained for diagnosis and treatment, in collaboration with doctors and hospitals. “Everything starts from the field” , stresses Nesma Houmani. “We work with the medical teams on their needs and data in order to refine the detection of pathologies, assess their severity and establish detailed patient phenotypes.”
A different vision of data management
Ultimately, algorithms can merge data and identify relationships between clinical variables that might otherwise escape the notice of medical teams. This would help to better characterize individuals and offer appropriate therapeutic pathways and treatments that best suit patients.
For now, Nesma Houmani’s team aims to exploit AI in the case of liver transplantation. The algorithms that are developed help establish the most relevant sets of variables and, above all, identify correlates between these variables in order to accurately predict patients’ risk of mortality. The team also seeks to integrate AI into a software used in walking rehabilitation centers. This tool allows researchers to analyze the quality of walking: is it symmetrical? To which extent is each joint affected? All this information can be analyzed to measure the deficiency and thus help assess the impact of a treatment.
At the crossroads of technological innovation, interdisciplinary scientific research and overall social development, Nesma Houmani’s work presents ‘a different vision of how to manage human-generated data’. “As an engineer by training, I need concrete cases, I need to know the purpose of what I’m doing. And the use of algorithms that take into account the constraints in the field and whose outcomes can be explained helps to do just that.”
Source : ip-paris / Nesma Houmani
Send us your press releases to [email protected]
Discover more from Impact AI News
Subscribe to get the latest posts sent to your email.
Jan 17, 2025 – By analyzing human-generated data, artificial intelligence…
