Foto del docente

Mauro Ursino

Full Professor

Department of Electrical, Electronic, and Information Engineering "Guglielmo Marconi"

Academic discipline: ING-INF/06 Electronic and Informatics Bioengineering

Director of First Cycle Degree of Biomedical Engineering

Research

1) Development of neural networks to simulate the integration of different sensory modalities (visual, auditory, tactile) in brain regions, in agreement with data from the neurophysiological literature; 2) Development of new methods for the estimation of effective brain connectivity among brain regions during cognitive and/or motor tasks, with the use of mathematical models of interconnected neural masses; 3) Development of models of neural oscillators to deal with the problem of different objectds representation in memory, object recognition from incomplete information, and link between objects and language. 4) Development of mathematical models able to describe the integrated action of different cardiovascular and cerebrovascular regulatory mechanims in the short period. 5) Application of advanced methods of signal processing (wavelets, Independent Component analysis) for the study of EEG signals and oculomotor signals in the awake state and/or during sleep, in normal and pathological subjects.

1) It is well known the existence in the brain of neurons which integrate information from different sensory modalities, such visual, auditory and tactile stimuli. This integration seems to play a major role in orienting overt behaviour (for instance movements) based of information coming from the external world. However, the mechanisms governing the working of these neurons are still insufficiently understood. Aim of this research is to develop physiologically realistic neural networks able to explain the main results in the neurophysiologic literature and the main psychophysical tests concerning multisensory integration. Subsequently, in cooperation with the Center for Cognitive Neurosciences (University of Bologna, Cesena, Prof. Ladavas) and with the Wake Forest University School of Medicine (North Carolina, USA, Prof. Stein) the models will be used to design new experiments, able to discriminate among different hypotheses.  The final aim is to arrive at a deeper theoretical understanding of the neural mechanisms which consent the integration of different sensory information, and their use in the clinical practice to improve cognitive deficits. 2) It is well known that the fulfilment of cognitive and motor tasks is the results of the participation of different brain regions mutually interconnected. Brain connectivity can be estimated starting from neuroimaging measurement, such as the functional magnetic resonance, the positron emission tomography, the electroencephalography and the magnetoencephalography. The methods usually adopted, however, are based on empirical models of the data. Aim of the research is to formulate new methods for the study and the analysis of brain connectivity, by using physiologically inspired mathematical models. To this aim, the so-called “neural mass models” can be particularly useful. In these models, the activity of entire populations of neurons is described using a few state variables, by neglecting individual fluctuations. The research (performed in cooperation with the Department of Human Physiology, University of Rome, Prof. F. Babiloni) is aimed at understanding the effect of connectivity among different brain regions on the quantities measured with non-invasive neuroimaging techniques and on the generation of EEG rhythms. In perspective, the methods can be adopted to design new and reliable non-linear methods to estimate brain connectivity from real data in individual subjects. 3) The representation of objects in brain memory occurs in a distributed fashion: an object is generally represented as a collection of features, each mapped in a different brain region. It is thus of great value to understand how features of the same object are linked together to form a unitary object representation (binding problem) and how features of different objects, simultaneously in memory, are separated (segmentation), in order to arrive at a coherent representation of the world.  A credited hypothesis assumes that object representation is achieved by the brain via the synchronization of neural groups, which code for different attributes of the object in different brain regions. Aim of the research is to develop mathematical models, based on the use of oscillating neural groups, to analyse and solve the binding and segmentation problem, and to reconstruct objects even starting from incomplete knowledge, both for what concerns a low-level representation (such segmentation of a visual scene) and for what concerns a high-level representation (recognition of abstract objects and its relation with language and semantics). A peculiarity of these models is the use of a realistic neural architecture and the use of physiological learning rules, based on physiological knowledge. 4) Maintenance of cardiovascular homeostasis is warranted by the action of sophisticate regulatory mechanisms, which operate at different levels in response to acute perturbations (such as hypoxia, hypercapnia, hemorrhage, etc...). Aim of the research is to develop and improve comprehensive mathematical models, which describe the main regulatory actions both at a systemic level (cardiac contractility, heart rate, circulation in less vital peripheral organs) and at the level of vital organs (coronary and brain circulation, exercising muscles). These models, developed and implemented with the cooperation of Medical Philips North America (Ing. Chbat) will be used as a tool for improving basic physiological knowledge, as a support for the clinical practice (for instance through estimation of clinical parameters, taking individual variability into account) and within the medical education framework. 5) The analysis of biomedical signals requires the use of advanced processing techniques, such as the wavelets and the Independent Component Analysis (ICA). Wavelet analysis is particularly useful to study non-stationary biomedical signals, in cases where the most important phenomena are characterized by transient events. ICA allows the search of the sources of a signal by excluding or reducing the effect of possible artefacts. These techniques are applied to two different biomedical signals: i) the scalp electroencephalogram (EEG) obtained in healthy and pathological subjects to detect and classify different patterns of cortical activity; ii) the electro-oculogram obtained on volunteers and pathological subjects (in cooperation with the Department of Neurogical Sciences, University of Bologna, Prof. Montagna) during wake and sleep periods. In particular, wavelets are used to detect the presence of slow-eye-movements, considered as a marker of neural structures involved in sleep regulation.