Objective evaluation of speech therapies for multiple sclerosis patients - An approach based on vocal indexes obtained with different software applications.

Multiple Sclerosis (MS) is a chronic demyelinating autoimmune disease that affects the central nervous system (CNS) causing a wide spectrum of signs and symptoms: motor symptoms such as spasticity and weakness, fatigue, dysphonia, speech or vocal disorders. Hypophonia, which is characterized by a reduction in the intensity of the voice, has a particular relevance among all the signs of the disease which demands for vocal analysis and therapies that allow to reduce social life limitations of MS patients. One of the most advanced hospital facilities that is involved in the MS research is the therapy and rehabilitation department of Don Carlo Gnocchi Foundation in Milan. This foundation deals with speech therapy and among these the Lee Silverman Voice Treatment (LSVT) therapy acquires particular relevance. This therapy is administred in an innovative way in telerehabilitation to ten MS subjects, allowing a comparison with the classic LSVT therapy administred in clinic to other ten patients. Therefore, the aim of the analysis conducted in this thesis is to evaluate the non inferiority of the telerehabilitation therapy compared to the in-clinic one, opening the possibility of using a therapy without problems of accessibility, waiting times and costs. This analysis is accomplished analyzing the speech material composed of three repetitions of a sustained vowel /a/, a free speech and a reading recording. These recordings are acquired with a vocal recorder equipped with an in-air microphone and with a contact microphone (Vocal Holter, VH) before the therapy (T0), after the therapy (T1) and after three months from the end of the therapy (T2). Since T2 recordings are not available for a sufficient number of patients, the analysis is focused entirely on sounds at T0 and T1. The evaluation is conducted extracting from these recordings the most significant vocal parameters that are Jitter, Shimmer, HNR, CPPS, F0 and SPL and computing the vocal indexes Acoustic Voice Quality Index (AVQI) and Warning Score (WS). This is performed using software commonly employed in the vocal analysis that are Praat, VOXplot and Matlab, analyzing their pre-processing, parameters extraction criteria and output. The analysis reveals similarities between Praat and VOXplot in terms of AVQImean±standard deviation, respectively returning values of 6.0±0.29 and 6.0±0.30 at T0. Also Matlab provides a similar output, resulting in an AVQI of 5.5±0.26. These AVQI values are compared to those obtained at T1, that are 5.2±0.29, 5.1±0.29 and 4.9±0.27 respectively for Praat, VOXplot and Matlab: this confirms a slight improvement of the disease after the therapy, considering that a lower AVQI corresponds to a healthier subject. In addition, the obtained results confirm the effectiveness and the reliability of the telerehabilitation therapy compared to the in-clinic one. Considering as example the delta AVQI (difference between T0 and T1) in Praat, the value of 0.5±0.19 for the telerehabilitation group is similar to the one obtained for the in-clinic group and equal to 0.9±0.20. The same analysis conducted for the AVQI is done in terms of WS, confirming also in this case the improvement of the disease for some patients as well as the effectiveness of the telerehabilitation. Finally, a comparison in terms of WS is conducted between the two types of microphones used for the acquisitions of the recordings.