This project aims to explore the links between olfaction and consciousness by using olfactory stimulation to assess the state of neurologically impaired patients. The method, combining neuroscience and generative models, could revolutionize medical care and diagnosis, with potential implications for diseases such as Parkinson’s disease.
Per Fumum Endowment Fund: What are the main objectives of your project on olfactory stimulation in coma patients?
Pierre-Marie Lledo: “Quite simply, we aim to take advantage of a unique feature of sensory organization in humans: olfaction is directly connected to limbic structures, which trigger responses beyond voluntary control and consciousness. Unlike other sensory modalities that pass through a ‘sorting gate,’ the thalamus, the sense of smell has direct access to emotional structures. For example, before even recognizing a wine, we instinctively know whether we like it or not. This specificity led us to explore olfaction, in addition to hearing, to analyze patients’ brain reactivity through different sensory pathways.”
Jean-Baptiste Masson: “As a physicist, we seek to leverage this specificity to ‘perturb’ patients in the scientific sense of the term—that is, to generate a measurable and natural signal that alters the patients’ brain characteristics. Through these perturbations, we hope to extract valuable information about their brain state using electroencephalograms (EEG).”
Per Fumum Endowment Fund: How can olfactory stimulation help assess the state of consciousness in brain-injured patients?
P.M.L.: “We combine exposure to odors with EEG measurements to observe patient reactivity, particularly in those on artificial ventilation. By measuring brain dynamics before and after odor exposure, we hope to determine whether a patient is on a recovery trajectory. This continuous monitoring will allow us to understand which variations indicate a potential for rehabilitation.”
J.B.M.: “We use a protocol integrating external and clinical data to identify patterns that indicate probable trajectories. The goal is to work probabilistically to assess whether an individual is following a favorable recovery path.”
Per Fumum Endowment Fund: What are the main challenges in integrating generative machine learning models with EEG measurements?
J.B.M.: “Generative models, like those inspired by large language models, require vast amounts of data. While we have a significant amount of data, it remains limited compared to textual corpora. Our challenge is to develop models that are both mathematically realistic and clinically validated. Mathematical realism ensures coherent properties, while medical validation guarantees that the generated signals are useful.”
Per Fumum Endowment Fund: What impact do you hope this project will have on medical research and therapeutic decisions for patients with severe brain injuries?
Jean-Baptiste Masson: “Our initial goal is to provide doctors with tools to enhance their decision-making process. These tools will never replace their expertise but will offer additional insights, such as trajectory indicators based on quantitative data. In the long run, we hope to generate new medical knowledge that will improve our understanding and prediction of patients’ responses to various treatments.”
Pierre-Marie Lledo: “By analyzing patient trajectories, we aim to go beyond a binary vision (awakening or not) and identify specific profiles, such as responders to certain treatments. This work could revolutionize how we approach diagnostics and therapies.”
Per Fumum Endowment Fund: Can you share any preliminary results or initial observations from your study conducted on 500 patients?
P.M.L.: “We have not yet started analyzing the data. The preprocessing phase is underway, but we prefer to wait before commenting.”
Per Fumum Endowment Fund: What are the next steps in your research, and what collaborations do you plan to pursue to advance in this field?
P.M.L.: “Key milestones we have recently achieved include recruiting suitable candidates. On the clinical side, the head of the intensive care unit at GHU Sainte-Anne has joined the laboratory for his thesis, and another PhD candidate will focus on the mathematical analysis of signals. We have also obtained all the necessary administrative approvals, including that of the ethics committee.”
J.B.M.: “On the mathematical side, we are working with a talented student who will join the project after completing his master’s degree at UCL in London. He will develop methods to encode signals using both existing and synthetic data. Clinically, our data comes exclusively from Sainte-Anne Hospital, where we have established a compliant and secure infrastructure. This required significant groundwork to ensure the trust of medical and administrative teams.”
