Andrés Gómez-Rodellar

Neurodegenerative motor disorders affect the neuromuscular system challenging daily life and normal activity. Parkinson’s Disease (PD) is among the most prevalent ones, with a large impact and rising prevalence rates. Speech is most affected by PD as far as phonatory and articulatory performance is concerned. Neuromotor activity (NMA) alterations have an impact on larynx muscles responsible for vocal fold adduction and abduction, hampering phonation stability and regularity. The main muscular articulators involved in phonation control are the cricothyroid (tensor) and thyroarytenoid (relaxer) systems, regulated by two distinct direct neuromotor pathways, activated by the precentral gyrus laryngeal control areas. These articulations control the musculus vocalis, directly responsible for regular vocal fold vibration. An indirect estimation of the muscular tension produced by inverse filtering may split into two independent channels, assumed to be the tensor and relaxer neuromotor pathways such as the differential neuromotor activity (DNMA). The amplitude distributions of both DNMA channels allow comparing phonations from PD-affected persons (PDPs) and age-matched healthy control participants (HCPs) with respect to a set of reference mid-age normative participants (RSPs). The comparisons are carried out by Jensen–Shannon distributions of PDP and HCP phonations with respect to those of RSPs. A dataset of 96 phonation samples from participants balanced by gender is used to train a set of decision tree classifiers (DTCs) to distinguish PDP from HCP phonation. The best results from 10-fold cross-validation offered accumulated mismatches of 0.09 and 0.1292 for male and female subsets. The sensitivity, specificity, and accuracy of the classification results when separating PDP from HCP phonatios were 93.33%, 88.23%, and 90.63% (male PDP versus HCP) and 92.86%, 83.33%, and 87.50% (female PDP versus HCP), providing a stratification of PDPs and HCPs by objective disease grading from explainable AI (XAI) methods.

Pathological voice characterization has received increasing attention over the last 20 years. Hundreds of studies have been published showing inventive approaches with very promising findings. Nevertheless, methodological issues might hamper performance assessment trustworthiness. This study reviews some critical aspects regarding data collection and processing, machine learning-oriented methods, and grounding analytical approaches, with a view to embedding developed clinical decision support tools into the diagnosis decision-making process. A set of 26 relevant studies published since 2010 was selected through critical selection criteria and evaluated. The model-driven (MD) or data-driven (DD) character of the selected approaches is deeply examined considering novelty, originality, statistical robustness, trustworthiness, and clinical relevance. It has been found that before 2020 most of the works examined were more aligned with MD approaches, whereas over the last two years a balanced proportion of DD and MD-based studies was found. A total of 15 studies presented MD characters, whereas seven were mainly DD-oriented, and four shared both profiles. Fifteen studies showed exploratory or prospective advanced statistical analysis. Eighteen included some statistical validation to avail claims. Twenty-two reported original work, whereas the remaining four were systematic reviews of others’ work. Clinical relevance and acceptability by voice specialists were found in 14 out of the 26 works commented on. Methodological issues such as detection and classification performance, training and generalization capability, explainability, preservation of semantic load, clinical acceptance, robustness, and development expenses have been identified as major issues in applying machine learning to clinical support systems. Other important aspects to be taken into consideration are trustworthiness, gender-balance issues, and statistical relevance.