Foto del docente

Elisa Magosso

Full Professor

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

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


Keywords: Short-term cardiorespiratory regulation Biomedical signal processing Mathematical models of physiological systems Neural networks Computational neuroscience Deep Learning

1) Modeling study of the cardiovascular, cerebrovascular and respiratory systems. The control mechanisms and their reciprocal interactions are analyzed by means of interpretative mathematical models.


2) Compartment models. Compartment models are developed to describe solute kinetics and fluid volume changes during renal dialysis, and toxin removal during liver dialysis.


3) Neural networks of sensory and cognitive functions. Neural network models are developed to study the integration among different sensory modalities (visual-acustic, visual-tactile), that takes place in some cortical regions. Neural oscillators models are realized to investigate the problems of object representation, learning, memorization, language, attention.


4) Advancede processing techniques applied to biomedical signals (time-scale analysis, principal component analysis, indipendent component analysis) for feature extraction, artifacts removal, source separation.


1) The cardiorespiratory and cerebrovascular systems are controlled by sophisticated regulation mechanisms that maintain arterial pressure, blood flows to tissues, oxygenation, etc. within tight limits despite the perturbations that daily affect these systems. The control mechanisms are generally studied separately in the literature and under experimental conditions that greatly differ from the actual ones (e.g, under artificial ventilation). A clear comprehension of the overall system and the interactions among the involved mechanisms is still lacking; moreover, because of the complexity of the relationships, the role of each single component within the overall integration cannot be identified using only clinical and experimental results. The proposed models aim at improving the knowledge of the cardiovascular and ventilatory regulation in different physio-pathological conditions: alteration of the inspired gas concentrations, anaemia, altitude, exercise, hypotension, cerebral ischemia, etc. These models, besides enhancing the physiological knowledge, may also be used as the core of didactic software for the education of technicians and physicians.


2) Complications during renal dialysis (such as hypovolemia and hypotension) still represent important clinical problems. Developing models of solute kinetics and fluid volume changes during kidney dialysis has important clinical implications. These model may be used i) in a predictive modality, to predict solute removal and blood volume changes during the dialysis session, starting from the initial patient state, thus foreseeing possible critical phases for the subject; ii) in a project modality, to set rationally the dialysate composition, in order to reach the end-dialysis targets minimizing the intradialitic risks.

The MARS (Molecular Adsorbent Recirculating System) technique is a clinically effective procedure for the temporary substitution of the liver function. This therapy is applied to patients with acute or chronic liver failure until the native liver function recovers or as a bridge treatment to liver transplantation. However, the exact mechanisms underlying the clinical effects of MARS therapy are poorly understood; this deficiency avoids the full exploitation of the MARS therapeutic potentialities and precludes the possibility of planning the optimal treatment on the basis of the specific conditions of the single patients. The research activity aims at developing models able to predict toxin removal during MARS session and to interpret the depurative mechanisms, identifying their specific role and temporal dynamics.


3) Neurons able to respond to stimuli of different modalities (visual, acoustic, tactile) have been identified in several cortical regions. Such integration seems to play a fundamental role both in orienting the behaviour (e.g., the gaze and movements) on the basis of the environmental information and to create a unitary and coherent perception of the external world. However, the neural mechanisms underlying this process are still largely unknown. The research activity is focused on the development of neural network models able to explain neurophysiological and psychophysical results related to multisensory integration, in order to achieve a deeper comprehension of the underlying neural mechanisms. Moreover, the realized models may be of value to investigate how the properties of the multisensory integration modify with development and experience.

Several cognitive functions result from the parallel activation of large populations of neurons, distributed among multiple and separated cortical regions. Hence, the comprehension of the mechanisms underlying the functional link among such distributed activity is of fundamental importance. A recent influential hypothesis sustains that this link is realized via the synchronization of the neural oscillatory activity (especially in gamma band, that is 30-100 Hz), that dynamically connect neurons into “functional webs”. Models of neural oscillators, based on physiologically plausible mechanisms, are of value to investigate and understand the synchronization and desynchronization mechanisms that occur in several cognitive processes such as object recognition, memory, learning, language, attention.


4) Interpretation of biomedical signals and extraction of information from them are highly complex tasks, both because of the presence of disturbances and noise contaminating the useful sifgnal, both because of the multiple components that that contribute to the signal origin. Advanced processing techniques (such as, discrete wavelet trasnform, principal component analysis, indipendent component analysis) are applied to biomedical signals in different conditions, in order to remove artefacts and noise, to extract specific events and features, to reconstrut signal sources. Specifically, these techniques are applied i) to scalp EEG tracings acquired in epileptic patients, in order to obtain information on the complex spatio-temporal dynamics of the seizure; ii)  to signals, such as the EEG and the electro-oculogram, acquired during poligraphic recordings, in order to identify specific events during the various phases of sleep and wakefulness, and analyze their relationships.