Voice analysis has significant potential in aiding healthcare professionals with detecting, diagnosing, and personalising treatment. It represents an objective and non-intrusive tool for supporting the detection and monitoring of specific pathologies. By calculating various acoustic features, voice analysis extracts valuable information to assess voice quality. The choice of these parameters is crucial for an accurate assessment. In this paper, we propose a lightweight acoustic parameter set, named HEAR, able to evaluate voice quality to assess mental health. In detail, this consists of jitter, spectral centroid, Mel-frequency cepstral coefficients, and their derivates. The choice of parameters for the proposed set was influenced by the explainable significance of each acoustic parameter in the voice production process. The reliability of the proposed acoustic set to detect the early symptoms of mental disorders was evaluated in an experimental phase. Voices of subjects suffering from different mental pathologies, selected from available databases, were analysed. The performance obtained from the HEAR features was compared with that obtained by analysing features selected from toolkits widely used in the literature, as with those obtained using learned procedures. The best performance in terms of MAE and RMSE was achieved for the detection of depression (5.32 and 6.24 respectively). For the detection of psychogenic dysphonia and anxiety, the highest accuracy rates were about 75 % and 97 %, respectively. The comparative evaluation was carried out to assess the performance of the proposed approach, demonstrating a reliable capability to highlight affective physiological alterations of voice quality due to the considered mental disorders.

As explainability seems to be the driver for a wiser adoption of Artificial Intelligence in healthcare and in critical applications, in general, a comprehensive study of this field is far from being completed. On one hand, a final definition and theoretical measurements of explainability have not been assessed, yet, on the other hand, some tools and frameworks for the practical evaluation of this feature are now present. This paper aims to present a concrete experience in using some of these explainability-related techniques in the problem of predicting the size of amino acids in real-world protein structures. In particular, the feature importance calculation embedded in Random Forest (RF) training is compared with the results of the Eli-5 tool applied to the Neural Network (NN) model. Both the predictors are trained on the same dataset, which is extracted from Protein Data Bank (PDB), considering 446 myoglobins structures and process it with several tools to implement a geometrical model and perform analyses on it. The comparison between the two models draws different conclusions about the residues’ geometry and their biological properties.