The 10-year European flagship project, the Human Brain Project, as well as the American equivalent Brain Initiative, define the comprehensive understanding of the human brain as a main challenge for the new millenium. A challenge that can be addressed only with a multidisciplinary approach.
The neuroinformatics laboratory (NILab) rises as a collaboration between FBK and CIMeC in order to promote interdisciplinary research in cognitive neuroscience. Neuroinformatics stands at the intersection of neuroscience and information science, and it provides theories and technologies for managing, analyzing, and modeling data collected with modern instruments of brain inspection used within the neuroscience community.
Connection to Data Science
The neuroscience studies are moving from the experimental setting with tens individuals towards larger sample population. The new trends are demanding to design data analysis over thousands of individuals. According to this view several projects have been developed to enable the sharing of common large repositories of data, like the american Human Brain Project and the UK BioBank.
This new scenario is representing a great opportunity to exploit machine learning techniques. The large amount of data, order of tens of terabytes, is the premise for a more robust inference process but represents also an open challenge from the point of view of data analysis and research reproducibility.
Research Impact on Health
The design and the development of computational methods are concerned not only with the understanding of structure and function of brain. The characterization of the alterations of brain structure and functions are of great impact in the clinical practice. The open challenge is to disentangle the interindividual differences with respect to pathological alterations due to neurological diseases.
The Neuroinformatics Laboratory is actively collaborating with the Operative Unit of Neurosurgery and the Department of Pathological Anatomy of S. Chiara Hospital. Currently we support the process of neurosurgical intervention planning by providing the computational tools for the reconstruction of the brain structural connectivity and the subsequent dissection of neuroanatomical tracts of interest for the patient.
Key Projects and Results
An excerpt of recent results on (i) dissection of structural brain connectivity, (ii) inference from brain connectivity, and (iii) estimate of effective brain connectivity.
Dissection of Structural Brain Connectivity
Dissection of neuroanatomical tracts is an hard task for a machine, but even for a human it is not straightforward to select among millions of connections. We address both challenges, manual and automated in-vivo virtual dissection of white matter.
- White Matter Tract Segmentation as Multiple Linear Assignment Problems
N Sharmin, E Olivetti, P Avesani
Frontiers in Neuroscience 11, 754, 2018
- Tractome: a visual data mining tool for brain connectivity analysis
D Porro-Muñoz, E Olivetti, N Sharmin, TB Nguyen, E Garyfallidis, P Avesani
Data Mining and Knowledge Discovery 29 (5), 2015
Inference for Brain Connectivity
New computational methods allow the characterisation of brain connectivity at different levels and modalities. Inference from these complex structures is not straightforward because we need to find a correspondence between different representations. We address the challenge of inference when brain connectivity structures are not represented as volumetric images.
- Alignment of tractograms as graph matching
E Olivetti, N Sharmin, P Avesani
Frontiers in neuroscience 10, 554, 2016
- Differential Effects of Brain Disorders on Structural and Functional Connectivity
S Vega-Pons, E Olivetti, P Avesani, L Dodero, A Gozzi, A Bifone
Frontiers in neuroscience 10, 605, 2017
Estimate of Effective Brain Connectivity
Among the different types of brain connectivity, the directed functional coupling among the activations of brain regions is the most challenging. The lack of a neurophysiological model of effective connectivity is one reason of the complexity of the task. We address this challenge adopting a supervised learning approach driven by data.
- Supervised Estimation of Granger-Based Causality between Time Series
D Benozzo, E Olivetti, P Avesani
Frontiers in neuroinformatics 11, 2017
- Bayesian estimation of directed functional coupling from brain recordings
D Benozzo, P Jylänki, E Olivetti, P Avesani, MAJ Van Gerven
PloS one 12 (5), e0177359, 2017