French
English
Keywords: machine learning, heterogeneous data fusion, population graph learning, prognostic model, neuroimaging
Disorders of consciousness (DOC) represent a broad spectrum of conditions characterized by impaired awareness of oneself and one's surroundings, which can occur as a result of various brain injuries (e.g., head trauma) and include coma, vegetative state, and minimally conscious state.
Although each of these states has distinct clinical characteristics, the accurate diagnosis of DOC in the acute phase, as well as the prognosis for the patient's progress in the months following this episode, are major challenges for clinicians.
Multimodal brain imaging (e.g., MRI, PET) is useful for more accurately assessing residual brain function. However, merging the massive amount of information provided by these different imaging modalities, supplemented by the patient's clinical data, is a complex task.
As part of the IMAGINA research project, we have compiled a unique database of subjects in acute comatose states (70 examinations) using a hybrid machine that allows simultaneous acquisition of functional information via positron emission tomography (PET) and functional or structural information via magnetic resonance imaging (MRI).
The objective of this project is to develop an artificial intelligence-based prognostic model for patient outcomes at 3 months, using heterogeneous information from multimodal imaging data and clinical data.
This work will build on the encouraging results of an initial statistical learning model combining MRI and PET imaging data. Several avenues will be explored in the fields of deep learning and population graph learning. The candidate will have access to a unique database collected as part of the ANR IMAGINA project, aggregating examinations of comatose subjects in the acute phase (>70 examinations) performed on a hybrid machine allowing the simultaneous acquisition of functional information in PET and functional or structural information in MRI.
Please refer to the attached file for details regarding the subject and application procedures.