Analisi del segnale electromiografico di superficie per diverse applicazioni e per la stratificazione di pazienti: applicazione in diverse patologie (Diabete, Sindrome dell'X Fragile, SMA)

The analysis of sEMG signals provides quantitative information regarding the amplitude, spectral features and timing of muscle activations, however there are both educational and technical barriers limiting its clinical use. There is a need to define a limited number of sEMG parameters that are applicable to specific pathologies of interest. In this context artificial intelligence methods have been recently adopted as an aid in identifying possible biomarkers of specific neuromuscular pathologies from biomechanical data. The purpose of this PhD project was to identify which parameters can be extracted from sEMG signals to characterize different pathologies affecting the musculoskeletal system, furthermore these parameters were used to classify the pathological subjects from healthy peers to verify the possibility to use sEMG signals to drive assisted support systems for clinical decision making. In particular the sEMG was extracted by adopting a wide variety of algorithms both in the temporal and frequency domain from three populations of subjects: diabetic subjects with and without neuropathy, children with Fragile X Syndrome (FXS) and Spinal Muscular Atrophy (SMA). In the first two pathologies the sEMG signal was extracted and analysed during gait while in the last one, similar algorithms were applied to signals derived during isometric masticatory muscle contractions. Regarding the application of sEMG analysis to diabetes, the temporal sEMG features were analysed to characterize the motor alterations associated with diabetes. With this purpose, unsupervised hierarchical clustering was applied to different combinations of the extracted sEMG feature. Results suggest that the muscle control alterations in diabetic subjects are too complex to be detected by a limited set of temporal parameters. Furthermore the presence of alterations caused by diabetes in the spectral parameters of sEMG signal was investigated. An increased instantaneous mean frequency value may be considered a biomarker of the motor control alterations present in subjects with diabetic neuropathy. For what concerns application of sEMG analysis to FXS, feasibility and repeatability of gait analysis in these subjects was confirmed. The temporal sEMG parameters were further investigated to identify the potential biomarkers by means of unsupervised classification. Co-contraction parameters proved to be able to stratify FXS and control subjects, as well as different FXS mutational categories. As a next step, seven different supervised learning algorithms were used with three different input vectors to classify four groups of subjects. The sufficient accuracy of the classification of FXS and control, as well as of different FXS mutational categories was possible only using sEMG parameters as the input for supervised learning algorithm. Finally, sEMG signals acquired in FXS subjects during gait for the first time were elaborated in the time-frequency domain. Visible differences were identified in the spectral attributes of Rectus Femoris and Biceps Femoris. The possibility to adopt the sEMG elaboration techniques previously used to investigate the function of lower limb muscles, to assess the masticatory muscles function was investigated. Preliminary results confirmed that the sEMG analysis techniques developed within this thesis could provide meaningful information also in regards to the function of masticatory muscles. During the doctoral project several algorithms to estimate different temporal and time-frequency parameters were developed and used to elaborate sEMG signals of healthy subjects and individuals affected by three different pathologies. For each of them clinically useful information were extracted. In conclusion, the aim of this thesis to identify which sEMG parameter can characterize the different pathologies and to use them to drive assisted support systems for clinical decision making was successfully accomplished.