Davide Orsenigo
PhD Student @ Università degli Studi di Torino
PhD student in Neuroscience, studying the brain through the lens of higher-order interactions and information theory.
Papers 6
-
Lorenz, G. M., Engel, N. M., Koçillari, L., Celotto, M., Orsenigo, D., Curreli, S., Blanco Malerba, S., Engel, A. K., Kayser, C., Fellin, T., Luppi, A. I., Panzeri, S.
Partial information decomposition (PID) has emerged as a principled way to decompose the information carried by neural activity into components identifying whether interactions among neurons or brain areas generate synergistic or redundant information. Here, we demonstrate that empirical measures of synergy and redundancy based on either Gaussian or discrete probability estimators suffer from a substantial limited-sampling estimation bias. This bias is much larger for synergy than for redundancy. The gap between them increases with the number of parameters specifying the probability distributions. We develop procedures that effectively correct for the bias and provide rules of thumb for the sample sizes required to obtain unbiased estimates. We show that, when used on empirical brain datasets, they successfully remove large synergy biases across species, recording modalities, and experimental designs. Our bias corrections extend the range of neuroscience questions and experimental designs addressable with PID and allow accurate comparisons between synergy and redundancy.
-
Dohnány, S., Jerotic, K., Orsenigo, D., Serra, E., Ali, H., Buhler, J., Liu, Z.-Q., Muta, K., Hata, J., Okano, H., Deco, G., Kringelbach, M. L., Luppi, A. I.
How the activity and connectivity of the brain support consciousness remains a central question in neuroscience. Recent progress driven by the use of functional MRI has seen growing recognition that large-scale distributed functional organisation of the human and non-human primate brain are systematically and consistently reshaped by anaesthetic-induced unconsciousness, across anaesthetics and across human and macaque. Here, we generalise these results to a different primate species that is gaining traction as model organism in neuroscience, the marmoset ( Callithrix jacchus ). We also generalise results to an additional anaesthetic, isoflurane, which we compare with propofol and sevoflurane. We report that under anaesthesia with propofol, sevoflurane, or isoflurane, distributed brain activity from functional MRI is increasingly constrained by the underlying structural connectivity across scales. Anaesthesia also induces a collapse of the principal gradient and intrinsic functional geometry of the marmoset brain, coinciding with a breakdown of hierarchical integration. Altogether, the present results indicate generalisable signatures of anaesthesia in the large-scale organisation of the primate brain.
-
Orsenigo, D., Luppi, A. I., Diano, M., Ciorli, T., Borriero, A., Willis, H. E., Petri, G., Bridge, H., Tamietto, M.
Damage to the primary visual cortex causes loss of conscious vision, yet some patients retain the ability to respond to stimuli despite reporting no visual experience. Why similar lesions produce such different behavioral phenotypes remains unclear. While research to date has focused primarily on spared pathways that bypass V1, here we asked whether these divergent outcomes are also linked to the brain's intrinsic functional architecture. In the largest resting-state fMRI cohort of patients with unilateral V1 damage reported to date, we quantified information sharing between regions across cortical and subcortical parcels in blindsight-positive and blindsight-negative patients, as well as in age-matched healthy controls. Despite comparable lesions, the two patient groups displayed distinct hierarchical patterns on the cortex: B+ patients preserved a sensory-to-association organization as in healthy controls, whereas B- patients exhibited a marked flattening of this hierarchy. The effect was driven by abnormally low shared-information coupling within unimodal cortices and scaled continuously with single-subject behavioral blind-field detection performance. A thalamic region consistent with the pulvinar, linking the contralesional visual cortex and the frontal eye field, discriminated B+ from B- patients. These findings highlight the system-level consequences of V1 damage supporting blindsight, suggesting that the unimodal-transmodal axis might track not only global states of consciousness, but also whether sensory information can guide behavior without awareness.
-
Santoro, A., Neri, M., Poetto, S., Orsenigo, D., Diano, M., Gatica, M., Petri, G.
We present the first large-scale comparison of higher-order interaction (HOI) metrics — spanning information theory and topology — applied to resting-state and task fMRI data from 100 Human Connectome Project subjects. We identify three distinct classes of HOI metrics: redundant, synergistic, and topological, with the latter bridging the two. Despite their differences, all metrics align with the brain's unimodal-to-transmodal hierarchy and, in some cases, with receptor architecture. HOI metrics outperform classical functional connectivity in fingerprinting and are more predictive of behavior, positioning them as key tools for linking brain architecture and cognition.
-
Orsenigo, D., Setti, F., Pagani, M., Petri, G., Luppi, A., Tamietto, M., Ricciardi, E.
We show that while the brain's large-scale functional architecture is shaped by innate hierarchical gradients, sensory deprivation induces targeted, experience-driven reorganization that flexibly reconfigures cortical connectivity without disrupting the brain's core scaffold. This work demonstrates how congenital sensory deprivation leads to specific adaptations in brain networks while preserving fundamental organizational principles.
-
Celeghin, A., Borriero, A., Orsenigo, D., Diano, M., Méndez Guerrero, C. A., Perotti, A., Petri, G., Tamietto, M.
This paper examines Convolutional Neural Networks (CNNs) as in silico models of the primate visual system, highlighting structural and functional parallels. It discusses challenges in modeling visual processing pathways and proposes architectural constraints to align CNNs more closely with biological vision, extending their applicability beyond object recognition.