This is the challenge addressed by a multicenter European study coordinated by Jacobo Sitt (Inserm) at Paris Brain Institute, with which Dragana Manasova was involved during her doctoral work. The researchers propose a novel approach: combining six assessment techniques, each capturing a distinct aspect of brain function: high-density electroencephalography (EEG) at rest and during an auditory task, structural and functional MRI, diffusion MRI, and positron emission tomography (PET). The resulting data are then processed using machine-learning algorithms to make them interpretable. The goal? To determine what each of these techniques reveals about consciousness and how they can be used to improve diagnosis and prognosis.
Multimodal Analysis Enhances Predictions of Patient Outcomes in Brain Injury
As part of the European PerBrain consortium, the team recruited nearly 400 patients across France, Germany, and Italy and compared their clinical outcomes with predictions from multimodal analysis. Their results indicate that combining data from multiple techniques significantly improves model performance: the more modalities available, the more reliable the predictions. The researchers also show that the modalities that contribute most to diagnosis are not necessarily the same as those that predict patient outcomes.
Functional vs. Structural Measures Inform Different Aspects of Prognosis
Functional measures of brain activity—metabolic measures from PET or electrical measures from EEG—provide useful information about patients’ current state of consciousness but are less informative about their evolution. Conversely, structural measures, such as diffusion MRI, which analyzes the integrity of brain connections, or conventional MRI, which assesses anatomical integrity, are more relevant for prognosis.
Discrepancies Between Modalities May Reveal Hidden Consciousness
Finally, discrepancies between modalities—for example, between EEG and MRI—are particularly common in patients who ultimately improve. In other words, these divergences, far from being problematic, may actually signal the presence of “islands of consciousness” that are not always captured through clinical observation alone.
“The aim of our study was to bring together a wide range of clinical and brain imaging data within a single, coherent analytical framework. By combining these rich and complementary sources of information, we sought to gain a better understanding of complex brain states, as they occur in real-world clinical practice. This work also provides insight into how computational analyses, including artificial intelligence models, can support medical decision-making and help clinicians make more informed choices,” says Dragana Manasova.
Multimodal Tool Aims to Standardize Consciousness Assessment Across Clinics
“Now that we have demonstrated the power and usefulness of multimodal analysis in assisting clinicians, we would like to see this tool adopted in expert centers,” explains Jacobo Sitt, co-head of the PICNIC Lab at Paris Brain Institute. “Clinical assessment of patients with disorders of consciousness is not conducted in the same way everywhere; it can vary depending on countries, professional cultures, or access to advanced technologies. Our aim is for all clinicians to share the same reference framework and be able to produce comparable data to advance consciousness research.”
Easy-to-Use Device Supports Personalized Patient Care
The tool developed by the team is a small, easy-to-use device suitable for clinical settings. It provides a probabilistic and integrative assessment of the patient’s condition, giving the medical team full latitude to contextualize and interpret the results. “This tool does not replace human expertise but offers a way to objectify often ambiguous clinical observations and personalize patient care with a view to achieving the best possible recovery,” the researcher concludes. “It also allows us to better understand the link between brain biology and subjective experience.”
Source: Paris Brain Institute