The work performed can be organized according to the three specific objectives of the project:
1) Implement whole brain models of the brain activity of each single patient and extract from these models physiologically relevant parameters and test them as putative biomarkers of the patients’ state of consciousness (MBMs):
- The successful implementation of subject modeling and the extraction of MBM from the complete cohort. Whole-brain model compounded by coupled non-linear oscillators. Different model parameter's optimization strategies were developed to fit the model to empirical data. (Fig1)
2) Benchmark the capacity of an automated machine-learning classifier combining these two types of biomarkers; testing whether synergies between EBMs and MBMs can be exploited and better prediction accuracy is obtained compared with the use of EBMs only.
- We obtained a combined set of model-based biomarkers (MBMs) and empirical biomarkers (EMBs) along the selected data set. The MBMs were based not only in the model parameters but also on markers derived from the inclusion of the non-equilibrium dynamics quantification of different states of consciousness. (Fig. 2)
- We are still working to find the best combination of EEG empirical biomarkers with fMRI model and empirical markers that better perform in classification tasks. In particular, we are progressing in dealing with the difficulties that carry the combination of modalities.
3) The objective is to use MBMs to better understand the physiopathology underlying disorders of consciousness and also to generate perturbations in silico that can provide insights into the prognosis of clinical treatments.
We selected specific perturbations that represent a different and non-in vivo obtained source of information to provide insight into the diagnostic and prognosis of acute DOC patients. We developed the full framework applied to DOC patients using deep learning algorithms and the whole-brain models developed in the previous objectives. The result is that we created a prototype of computational open-source software that includes the generation of in silico perturbation that can be used as a clinical guideline.
Dissemination activities
3 international conferences in the poster sessions: The association for the scientific study of consciousness (ASSC) conference (July 2023, New York, USA), Organization for Human Brain Mapping, a general (OHBM) conference (July 2023, Montreal, Canada) and Barcelona Computational, Cognitive and Systems Neuroscience Community (Barccsyn) conference (May 2023, Barcelona, Spain). I was also invited as a speaker in 2 international Workshops: IBD, the interpretable Brain Data Workshop (June 2023, Stockholm, Sweden), and Bernstein Computational Neuroscience Satellite Workshop: Whole brain dynamics: Modeling and applications (Sep 2023, Berlin, Germany). (This last workshop was developed after the two-year project, but I was invited before that time and as the results obtained during these two years). I also disseminated and communicated project activities and results via Twitter from my profile: @ysanz6
Outreach activities:
1) GIGA consciousness seminar. The seminar from GIGA consciousness group from University of Liege, Belgium. (Sept 2021, on line).
2) Computational Domain meeting. Meeting for the whole ICM focused on disseminating the computational approach that was developed in the institute. (Nov 2021, Paris, France)
3) Newspaper articles highlighting two investigations in collaboration with Dr. Deco
