RICERCA ITALIANA     

Research program

Computational and neural mechanisms of sensorimotor learning and control

University Co-ordinator

Università degli Studi ROMA TRE - ELETTRONICA APPLICATA - ROMA(RM)

Research Unit Leader

Tommaso D'ALESSIO

Description

The UR_UNIRoma3 will mainly deal with WP3, WP4, WP7, and WP8.

It will also collaborate with other URs to the definiton of specificatons for the virtual environemts and the definition of the experimental protocols.

Relative resource allocation: 10%.
Deliverable:
- D6.4: Report on the estimation of some neural correlates during the combined adaptation experiments (month 24; with all other URs)


WP3: NEURAL CORRELATES OF MOTOR LEARNING

In collaboration with UR_UNIGE_2, the neurology unit of UR_UNIRoma3 will contribute to the definition of the stimolation modalities in terms of quality, quantity, amplitude and timings, with respect to the cortical area involved and the motor task to be executed.
In particular, UR_UNIRoma3 will deal with the use of repetitive Transcranial Magnetic Stimulation (rTMS) to investigate synaptic plasticity of excitatory and inhibitory circuits in a group of subjects undergoing tasks defined in the previous WPs. rTMS will be applied to differentiate Short Term Potentiation (STP) effect, from Long Term Potentiation (LTP), considered as a marker for synaptic plasticity, with different duration on the amplitude of motor evoked potentials (MEP) and cortical silent period (CSP) obtained after the end of stimulation at different intervals.
For the execution of the trials, UR_UNIRoma3 is equipped with a High Frequency Magnetic Stimulator (Magstim Rapid, The Magstim Company Ltd, Whitland, South West Wales, UK), that comes up with the specifications of the stimultion trials.
Data collected in this WP will help verify if rTMS is able to provide information on the mechnisms of neural plasticity, cosidered as the physiological foundation of learning/adaptation mechanisms. Moreover, by this pilot study, it will be possible to check if rTMS can play a role in the assessment and follow-up of changes in cortical excitability in patients with impairment of learning and memory.

Relative resource allocation: 15%
Deliverables:
- D3.3: Report on experiments performed with TMS (month 15; with UR_UniGE2)
- D3.4: Report on experiments performed with rTMS (month 14)

WP4 - ELECTRICAL MUSCLE ACTIVITY (EMG)

Motor learning in humans and primates is carried on by control mechanisms managed by Central Nervous System. The CNS modulates sensor inputs and builds correspondence maps to link movements in the end-effector space to neural outputs in the body space.
Correspondence maps are generally associated to the internal model hypothesis. Since the mechanism driving internal models construction is not completely known, techniques devoted to monitor model generation acquire a clear interest.
The electrical muscle activity (EMG) is widely considered as an indirect measurement of neural output and then can infer information on internal model generation. The acquisition of motor skills is connected to the appearance of the so called muscular synergies, that are those strategies adopted by the central nervous system to coordinate either motor unit groups or muscular complexes. The muscle coordination aims at obtaining functional assemblies devoted to guarantee a trade-off between task completion and control requirements simplification.
The study of muscular synergies construction can provide insights on motor learning mechanisms. In particular, that study should monitor if and how myoelectric correlates vary with respect to motor learning. Therefore, besides typical performance indexes, such as cinematic and dynamic parameters, electrical indexes extracted from muscular signals can be used to assess muscular synergies.
The analysis of motor learning mechanisms needs experimental protocols aiming at destructuring existing internal models by altering the environment. The alteration of the environment can be obtained by dynamical force fields (WP1) and by dynamical visual distortions (WP2).
Muscular synergies built during motor tasks execution in altered environment should represent a trade-off between environment management and controlling parameters reduction. When new motor skills are acquired neural output is altered and modulated by the task. Changing task requirements generates qualitative and quantitative modifications of both muscular patterns and segmental recruitment.
Even if the muscular synergies are typically defined as spatial patterns, on the basis of muscles used for task execution, the modifications of temporal patterns, i.e. timing of muscular activation, drive different behaviors of the end-effector. To execute a single motor task the number of muscular combinations is not so wide and it is determined by the force level each muscle can exert.
In order to monitor construction, modification and stabilization of muscular synergies on the basis of electrical indexes extracted by sEMG signals, suitable processing techniques need to be developed.
UR_UNIRoma3 is experienced in the development of processing techniques for sEMG signals acquired during static and dynamic protocols. sEMG signals acquired during motor learning protocols are characterized by high variability and then need to be processed by using non-stationary and statistical adaptive techniques.
In particular, it is important to provide techniques to detect the timing of muscular activations, to estimate signal amplitude to be correlated to the exerted muscular force, to evaluate frequency modifications driven by muscular fatigue phenomenon.
In order to detect the timing of muscular contractions, UR_UNIRoma3 will use the double-threshold statistical detector, already presented in (Bonato, D'Alessio, Knaflitz 1999) and widely used in clinical applications (Benedetti 1999). This detector has been widely tested in terms of noise robustness and evaluates the detection thresholds on the basis of the signal to noise ratio estimation. This approach should provide the monitoring of the timing modifications and the detection of muscular cocontraction-coactivation phenomena.
Some studies in the literature (Corcos et al., 1993; Gottlieb, 1994, 1996) refer to signal amplitude modifications as indexes to monitor neural output changes. However, the assessment of these indexes is quite difficult because of the signal amplitude variability and the non-analytical relation with the force generation phenomenon.
The Joint Analysis of EMG Spectrum and Amplitude (JASA) protocol, proposed by Luttmann and coworkers in 1996, tries to monitor force generation and fatigue condition on the basis of electrical indicators, i.e. amplitude and mean frequency of sEMG signal, extracted from myoelectric signals. UR_UNIRoma3 proposes to modify the protocol, with special attention to processing techniques, in order to code the muscular status using a four status classification: force generation, force decrease, muscular fatigue, recovery from muscular fatigue. Muscular statuses are detected by simultaneous modifications of the electrical indicators. In particular, the simultaneous increase of amplitude and mean frequency is coded as force generation, while the simultaneous decrease of both parameters is coded as force decrease; amplitude increase and frequency decrease is associated to a fatigue status, amplitude decrease and frequency increase is coded as recovery from fatigue.
Algorithms developed to estimate the electrical indicators, i.e. amplitude (D'Alessio e Conforto, 2001) and mean frequency (Conforto e D'Alessio, 1999), have been tested on non-stationary time series, are based on adaptive estimation techniques in frequency and time domains, and can be implemented in real-time. The information extracted from sEMG signals by these techniques is very robust in either static and dynamic conditions.
The coding of muscular status can greatly help for monitoring construction and modification of muscular synergies. Particular attention can be devoted to force conditions and muscular fatigue development which, especially when using varying force fields, can affect motor learning process.

Relative resource allocation: 35%
Deliverables:
- D4.1: First release of the modified JASA protocol (month 6; with UR_UniGE1)
- D4.2: Report on experiments on the construction of muscle synergies in the VHE (month 15; with UR_Polito,UR_UniGE1)
- D4.3: Report on experiments on the construction of muscle synergies in the VVE (month 12; with UR_Polito,UR_UniBO)
- D4.4: Report on experiments on muscle fatigue in the VHE (month 14; with UR_Polito,UR_UniGE1)

WP7 - COMPUTATIONAL MODELS

In this WP, UR_UNIRoma3 will coordinate the development of computational models for the extraction of primitives in the learning/adaptation process, from experimental data extracted throughout the project, in previous WPs. Among the possible models, performance indexes will be identified and chosen to obtain an objective and reliable evaluation of learning/adaptation mechanisms.
These performance indexes will nonetheless be used as input for the reduction of data dimensionality, based on redundancy check, and to distinguish data variability determined by the adaptation/learning process from other sources of data variability. To this objective, the computational methods that will be applied owe to the framework of standard multivariate analysis, such as clustering techniques, principal component analysis, kmeans. Moreover, descriptive methods based on fuzzy logic seem to meet the specifications of qualitative rather than quantitative modifications driven by adapting/learning a different/new task or evironment, and their application will be tested in terms of reliability.
The development of a computational model for the extraction af adaptation/learning markers will be accompanied by the development of functional/interpretative connectionist model for experimental data checking. This distributed model will mimic the activity of the set (neural system-biomechanical plant), and will be fed by inputs corresponding to the sensorial inputs given to the subject in the execution of the tasks.
The structure of the system will be composed of three different elements: a Feed-forward Artificial Neural Network, that will deal with the construction, reinforcement, and adaptation of the Internal Model corresponding to the experimental tasks; a Feed-back Neural Controller, that will simulate the on line modifications of the neural commands according to the changes of the sensorial inputs obtained during the execution of movements; the Biomechanical Model that will simulate the structure, at the macroscopic level, of the action exerted by the muscles, as driven by the neural commands.
The overall system will produce outputs dimensionally consistent with the experimental outputs obtained in the execution of the experimental trials.

Relative resource allocation: 30%
Deliverables:
- D7.1: Report on the multivariate analysis of data (month 18)
- D7.2: Report on the 3-element connectionist model (month 22)

WP8 - NEUROREHABILITATION

In the final WP UR_UNIRoma3 will carry out a pilot study on a small number of patients with mild cognitive impairment, a group of patients with a diagnosis of Alzheimer's disease at different stages (mild, moderate and severe) and a group of neurologically healthy age-matched control subjects. In the course of learning protocols, it will be possible to compare the variables reflecting cortical excitability, and thus to gain insights on mechanisms underlying motor learning/adaptation.
Data collected in this WP will help to verify if rTMS is able to provide information on the mechanisms of neural plasticity, thus allowing the assessment and follow-up of changes in cortical excitability in patients with impairment of learning and memory.

Relative resource allocation: 10%
Deliverables:
- D8.2: Report on rehabilitation experiments focused on patients with mild cognitive impairment (month 24)