Usefulness of acoustic studies on the differential diagnostics of organic and functional dysphonia

Evaluating the Diagnostic Potential of Connected Speech for Benign Laryngeal Disease Using Deep Learning Analysis

OBJECTIVES

This study aimed to evaluate the performance of artificial intelligence (AI) models using connected speech and vowel sounds in detecting benign laryngeal diseases.

STUDY DESIGN

Retrospective.

METHODS

Voice samples from 772 patients, including 502 with normal voices and 270 with vocal cord polyps, cysts, or nodules, were analyzed. We employed deep learning architectures, including convolutional neural networks (CNNs) and time series models, to process the speech data. The primary endpoint was the area under the receiver's operating characteristic curve for binary classification.

RESULTS

CNN models analyzing speech segments significantly outperformed those using vowel sounds in distinguishing patients with and without benign laryngeal diseases. The best-performing CNN model achieved areas under the receiver operating characteristic curve of 0.895 and 0.845 for speech and vowel sounds, respectively. Correlations between AI-generated disease probabilities and perceptual assessments were more pronounced in the connected-speech analyses. However, the time series models performed worse than the CNNs.

CONCLUSION

Connected speech analysis is more effective than traditional vowel sound analysis for the diagnosis of laryngeal voice disorders. This study highlights the potential of AI technologies in enhancing the diagnostic capabilities of speech, advocating further exploration, and validation in this field.

What is the functional/organic distinction actually doing in psychiatry and neurology?

The functional-organic distinction aims to distinguish symptoms, signs, and syndromes that can be explained by diagnosable biological changes, from those that cannot. The distinction is central to clinical practice and is a key organising principle in diagnostic systems. Following a pragmatist approach that examines meaning through use, we examine how the functional-organic distinction is deployed and conceptualised in psychiatry and neurology. We note that the conceptual scope of the terms 'functional' and 'organic' varies considerably by context. Techniques for differentially diagnosing 'functional' and 'organic' diverge in the strength of evidence they produce as a necessary function of the syndrome in question. Clinicians do not agree on the meaning of the terms and report using them strategically. The distinction often relies on an implied model of 'zero sum' causality and encourages classification of syndromes into discrete 'functional' and 'organic' versions. Although this clearly applies in some instances, this is often in contrast to our best scientific understanding of neuropsychiatric disorders as arising from a dynamic interaction between personal, social and neuropathological factors. We also note 'functional' and 'organic' have loaded social meanings, creating the potential for social disempowerment. Given this, we argue for a better understanding of how strategic simplification and complex scientific reality limit each other in neuropsychiatric thinking. We also note that the contribution of people who experience the interaction between 'functional' and 'organic' factors has rarely informed the validity of this distinction and the dilemmas arising from it, and we highlight this as a research priority.

Acoustical recognition of laryngeal pathology using the fundamental frequency and the first three formants of vowels

The recognition of laryngeal pathology by analysis of the voice is investigated. The fundamental frequency and the first three formants are considered. The recognition strategy is based on comparison with normal ranges calculated over 200 ordinary voices, grouped in ten age classes ranging from 20 to 70 years, for males and females. 220 test voices are studied divided into four groups: normal voices, functional dysphonia, nodules and recurrent nerve palsy. Each subject is marked according to his/her normal range. Parameters (or items) are calculated on the Interactive Laboratory System workstation. The vocalic material is composed of 11 vowels taken from a sentence. Results are given in terms of the number of values out of the normal ranges. Statistical analysis considers both parameter ability and error rates in pathology recognition. Pathology recognition shows the following error percentages: 23% for dysphonia, 14% for nodules and 33% for recurrent nerve palsy. Parameters do not show the same efficiency for voice pathology characterisation. Formants appear to be better than the fundamental frequency.