INRIA Grenoble PhD Seminars

February 2026

Speakers :

• Arturo Cabrera Vazquez, PhD student at STATIFY team (Inria) : Neuroinflammation and disrupted functional connectivity in coma: a graph-theoretical study
  Website Linkedin
  Coma, regardless of etiology, is a major health concern, often leading to lasting cognitive and behavioral impairments. Despite this impact, the structural and functional brain damage underlying coma remains poorly understood.¹ Recent TSPO PET studies have revealed substantial neuroinflammation in cortical and subcortical regions corresponding to key resting-state network nodes.² Resting-state functional connectivity (FC) breakdown in coma has been studied with graph-based measures, such as the Hub Disruption Index (HDI), revealing reorganization of functional networks.³ However, the relationship between neuroinflammation and FC in coma remains unclear.⁴
  To address this gap, we developed new modeling strategies to explore the relationship between neuroinflammation and FC in coma. Single-subject TSPO images were transformed into graphs by comparing within-subject regional neuroinflammatory profiles using the Bhattacharyya coefficient (BC). This process yields networks with nodes aligned across imaging modalities. Network reorganization within each modality was analysed using HDI. Cross-modal relationships were explored by introducing a multilayer notion of node equivalence⁵ that spans modalities and node metrics: nodes are characterized by their roles, defined by node metric values. Their roles are compared across layers of a multiplex network. In this framework, each layer represents relationships between brain regions in a specific modality (FC or TSPO). This approach enables the identification of shared cross-modal patterns, assessed using the Correspondence Structural Pattern Score (CSPS).⁵
  HDI results show that TSPO networks exhibit hub reorganization patterns that parallel those observed in fMRI data, suggesting that neuroinflammation mirrors aspects of functional network disruption in coma. Multilayer, fMRI and TSPO analyses further reveal a shared network structure across modalities that varies by condition, with mismatches between functional and neuroinflammatory hubs. Our findings point to whole-brain, cross-system reorganization in coma, underscoring the importance of unified multimodal frameworks, and suggesting new avenues for identifying mechanistic biomarkers.
1. Dehaene et al. (2021). Experimental and Theoretical Approaches to Conscious Processing. doi.org/10.1016/j.neuron.2011.03.018
2. Sarton et al. (2024). Neuroimmune activation is associated with neurological outcome in anoxic and traumatic coma. doi.org/10.1093/brain/awae045
3. Malagurski et al. (2019). Topological disintegration of resting state functional connectomes in coma. doi.org/10.1016/j.neuroimage.2019.03.012
4. Cabrera Vazquez et al. (2025). Graph theory in Disorders of Consciousness: Toward multimodal integration. https://hal.science/hal-05153617v1
5. Carboni et al. (2023). Nodal-statistics-based equivalence relation for graph collections. doi.org/10.1103/physreve.107.014302



• Luben Miguel Cruz Cabezas, visiting PhD candidate to STATIFY (University of São Carlos) : Conformal Prediction
  Website Google Scholar Linkedin
  Conformal Prediction (CP) is a rigorous distribution-free framework for uncertainty quantification, capable of providing marginally valid predictive inference for black-box models through prediction sets.
  Despite its robust guarantees, standard CP methods face significant challenges regarding local adaptivity, applicability to complex data settings, and the disentanglement of distinct sources of uncertainty.
  This presentation introduces the fundamentals of the CP framework and details our novel contributions designed to address these limitations. Furthermore, we demonstrate how we have expanded the scope of Conformal Prediction beyond standard prediction tasks.