Current standard protocols used in the clinic for diagnosing COVID-19 include molecular or antigen tests, generally complemented by a plain chest X-Ray. The combined analysis aims to reduce the significant number of false negatives of these tests and provide complementary evidence about the presence and severity of the disease. However, the procedure is not free of errors, and the interpretation of the chest X-Ray is only restricted to radiologists due to its complexity. With the long term goal to provide new evidence for the diagnosis, this paper presents an evaluation of different methods based on a deep neural network. These are the first steps to develop an automatic COVID-19 diagnosis tool using chest X-Ray images to differentiate between controls, pneumonia, or COVID-19 groups. The paper describes the process followed to train a Convolutional Neural Network with a dataset of more than 79, 500 X-Ray images compiled from different sources, including more than 8, 500 COVID-19 examples. Three different experiments following three preprocessing schemes are carried out to evaluate and compare the developed models. The aim is to evaluate how preprocessing the data affects the results and improves its explainability. Likewise, a critical analysis of different variability issues that might compromise the system and its effects is performed. With the employed methodology, a 91.5% classification accuracy is obtained, with an 87.4% average recall for the worst but most explainable experiment, which requires a previous automatic segmentation of the lung region.

Although a large amount of acoustic indicators have already been proposed in the literature to evaluate the hypokinetic dysarthria of people with Parkinson’s Disease, the goal of this work is to identify and interpret new reliable and complementary articulatory biomarkers that could be applied to predict/evaluate Parkinson’s Disease from a diadochokinetic test, contributing to the possibility of a further multidimensional analysis of the speech of parkinsonian patients. The new biomarkers proposed are based on the kinetic behaviour of the envelope trace, which is directly linked with the articulatory dysfunctions introduced by the disease since the early stages. The interest of these new articulatory indicators stands on their easiness of identification and interpretation, and their potential to be translated into computer based automatic methods to screen the disease from the speech. Throughout this paper, the accuracy provided by these acoustic kinetic biomarkers is compared with the one obtained with a baseline system based on speaker identification techniques. Results show accuracies around 85% that are in line with those obtained with the complex state of the art speaker recognition techniques, but with an easier physical interpretation, which open the possibility to be transferred to a clinical setting.

There exist many acoustic parameters employed for pathological assessment tasks, which have served as tools for clinicians to distinguish between normophonic and pathological voices. However, many of these parameters require an appropriate tuning in order to maximize its efficiency. In this work, a group of new and already proposed modulation spectrum metrics are optimized considering different time and frequency ranges pursuing the maximization of efficiency for the detection of pathological voices. The optimization of the metrics is performed simultaneously in two different voice databases in order to identify what tuning ranges produce a better generalization. The experiments were cross-validated so as to ensure the validity of the results. A third database is used to test the optimized metrics. In spite of some differences, results indicate that the behavior of the metrics in the optimization process follows similar tendencies for the tuning databases, confirming the generalization capabilities of the proposed modulation spectrum metrics. In addition, the tuning process reveals which bands of the modulation spectra have relevant information for each metric, which has a physical interpretation respecting the phonatory system. Efficiency values up to 90.6 \% are obtained in one tuning database while in the other the maximum efficiency reaches 71.1 \%. Obtained results also evidence a separability between normophonic and pathological states using the proposed metrics, which can be exploited for voice pathology detection or assessment.

Literature documents the impact of Parkinson’s Disease (PD) on speech but no study has analyzed in detail the importance of the distinct phonemic groups for the automatic identification of the disease. This study presents new approaches that are evaluated in three different corpora containing speakers suffering from PD with two main objectives: to investigate the influence of the different phonemic groups in the detection of PD and to propose more accurate detection schemes employing speech. The proposed methodology uses GMM-UBM classifiers combined with a technique introduced in this paper called phonemic grouping, that permits observation of the differences in accuracy depending on the manner of articulation. Cross-validation results reach accuracies between 85% and 94% with AUC ranging from 0.91 to 0.98, while cross-corpora trials yield accuracies between 75% and 82% with AUC between 0.84 and 0.95, depending on the corpus. This is the first work analyzing the generalization properties of the proposed approaches employing cross-corpora trials and reaching high accuracies. Among the different phonemic groups, results suggest that plosives, vowels and fricatives are the most relevant acoustic segments for the detection of PD with the proposed schemes. In addition, the use of text-dependent utterances leads to more consistent and accurate models.

Speech-based approaches for assessing Parkinson's Disease (PD) often rely on feature extraction for automatic classification or detection. While many studies prioritize accuracy by using non-interpretable embeddings from Deep Neural Networks, this work aims to explore the predictive capabilities and language robustness of both feature types in a systematic fashion. As interpretable features, prosodic, linguistic, and cognitive descriptors were adopted, while x-vectors, Wav2Vec 2.0, HuBERT, and TRILLsson representations were used as non-interpretable features. Mono-lingual, multi-lingual, and cross-lingual machine learning experiments were conducted leveraging six data sets comprising speech recordings from various languages: American English, Castilian Spanish, Colombian Spanish, Italian, German, and Czech. For interpretable feature-based models, the mean of the best F1-scores obtained from each language was 81% in mono-lingual, 81% in multi-lingual, and 71% in cross-lingual experiments. For non-interpretable feature-based models, instead, they were 85% in mono-lingual, 88% in multi-lingual, and 79% in cross-lingual experiments. Firstly, models based on non-interpretable features outperformed interpretable ones, especially in cross-lingual experiments. Specifically, TRILLsson provided the most stable and accurate results across tasks and data sets. Conversely, the two types of features adopted showed some level of language robustness in multi-lingual and cross-lingual experiments. Overall, these results suggest that interpretable feature-based models can be used by clinicians to evaluate the deterioration of the speech of patients with PD, while non-interpretable feature-based models can be leveraged to achieve higher detection accuracy.