Leader: Antonio Lanatà (UNIFI); Other collaborator(s):
In this task innovative wearable textile sensors for physiological monitoring along with intelligent platform integration will be implemented employing spatially distributed systems model to ensure security and reliability of the information. This task includes innovative signal processing and artificial intelligence tools for psychophysiological assessment and physical body activity analysis. Developed methods will enable the investigation of user engagement, cognitive loading, emotional response and mood dynamics to prevent physical and mental dangerous conditions. Furthermore, the dynamics of emotional face expression and eye gaze pattern will be processed for quantitative index extraction emotional wellness.
During the specified period (November 2023 - March 2024), our activity focused on different experimental designs for the psychophysiological characterization of young and adult subjects by analysing multivariate physiological signals obtained from biosensors and imaging techniques. In particular, a public database of EEG and fMRI data relating to an adult population considered at risk of developing Alzheimer Disease (AD) was preliminarily explored, along with a genotypic and phenotypic investigation. In experimental designs proposed to a healthy population, electrophysiological characterization of the participants was carried out through analysis of Event-Related Potentials (ERP) on words and pseudowords aloud reading task, a clinical protocol widely adopted by speech therapists and laryngologists to diagnose speech and language disorders. The ERP analysis was also proposed in an experimental design to understand subjects' implicit attitudes through the Implicit Association Test (IAT). In this experimental paradigm, eye gaze, blinking activity and variation of the pupillary diameter were investigated. Bibliographical research focused on rehabilitation and treatment aspects of elderly population with neurodegenerative disorders has been carried out. Particular attention has been paid to the study of the effect of non-invasive brain stimulation (tES and rTMS) in experimental contexts related to multi-sensory stimulation integrated into contexts of extended reality (XR) or serious games (SG). Finally, new acquisitions utilizing wearable sensors were performed on an experimental design to investigate possible different emotional and cognitive responses on real or AIgenerated human faces. Additional questionnaires to establish the participants' memory abilities and to understand the adopted strategy during the proposed task have been implemented.
Please see the next reporting period.
This report details activities conducted from July to September 2024, focusing on clustering analysis of vocal properties of 287 patients diagnosed with benign lesions of the vocal folds (BLVF) and unilateral vocal fold paralysis (UVFP), which are among the highest causes of dysphonia. The research aimed to identify if these patients constituted separate vocal subtypes of dysphonia and to understand whether misclustered data could depend on a specific diagnosis and age. This was done primarily to identify straightforward differences in vocal fold motor dynamics that generally are assessed with visual inspection through high- resolution endoscopy, which is not always available. Also, a particular emphasis was placed on misclustered observations to understand the role of confounding factors (i.e., the BLVF subtypes and age) in clustering analysis. Data collection regarded 287 patients recruited at the Ospedale Maggiore Policlinico di Milano. The diagnosis was made by perceptual voice evaluation (GRB scale) and video- laryngostroboscopic assessment. 136 patients presented UVFP and 151 BLVF. For the acoustic analysis, patients were asked to utter the sustained vowel /a/ for at least 3s. Acoustic parameters were extracted, accounting for age, gender and type of vocal emission. Specifically, fundamental frequency F0, local jitter, adaptive normalised noise energy, the first three formants, the duration of voiced and unvoiced parts, and the median, standard deviation, minimum and maximum values were used for further analysis. The k-means algorithm was applied to find clusters on the event space by iteratively assigning each point to the nearest candidate cluster centre and updating the cluster’s centres until convergence. The silhouette score offered a means for assessing the clustering quality by comparing inter-cluster distances with intra-cluster ones. Unaware and aware analyses were performed for clustering, whilst a misclustered data analysis was additionally performed to provide interpretable results for clinicians. Key findings for the female dataset in the unaware condition report that the best results were obtained with the principal component analysis (PCA). Also, a higher silhouette score for the UVFP class could be helpful in clinical practice for better recognising the most severe of the considered pathologies, possibly reducing its misdiagnosis. Results from the male dataset remain similar between the unaware and aware conditions. Moreover, in both genders, misclustered observations seem to be independent of a specific pathology (and its subtypes). Finally, the PCA weight analysis highlighted that phonation parameters were the most contributive. In conclusion, the results showed valuable preliminary data on voice disorders with clustering techniques on voice features, offering a new perspective for clinical practice and professionals.
During October-December 2024, we undertook several significant initiatives to advance research in physiological signal acquisition and analysis in clinical and experimental contexts. Below is a summary of the key activities conducted during this period:
These studies are relevant to ageing and healthy ageing, exploring the physiological underpinnings of emotional and social interactions. Emotional regulation and social interactions are critical for maintaining mental health and cognitive function in older adults, underscoring the broader implications of this research. Future efforts will further refine these methodologies and expand collaborations to amplify the impact of our findings.
Scientific publications