Measurement of communication impairment after treatment for oral and oropharyngeal cancer by automatic analyses of spontaneous speech associated with biopsychosocial factors - Archive ouverte HAL Access content directly
Conference Papers Year : 2023

Measurement of communication impairment after treatment for oral and oropharyngeal cancer by automatic analyses of spontaneous speech associated with biopsychosocial factors

Abstract

Background The perceptual evaluation of speech disorders produces scores that insufficiently predict the communication abilities in head and neck cancer, which may be mitigated by automatic speech analysis. The objective is to measure communication impairment using automatic analyses of spontaneous speech and biopsychosocial questionnaires in patients treated for oral/oropharyngeal cancer. Methods Automatic tools were applied to the spontaneous speech signal, to obtain scores relating to the voice, phonology, morphosyntax, semantics. Relevant and non-redundant parameters were retained by reduction of dimensionality. Ten self-questionnaires (assessing communication, associated deficits, communication needs relating to social circles, quality-of-life) were conducted. A predictive modeling of communication impairment (LASSO regression) was performed, using the scores from the automatic tools alone, which were then combined with the scores for the associated factors. Results Semi-structured interviews of 25 patients were recorded (median 67 years, IQR 12; oral cavity: 14, oropharynx: 10, both: 1). 149 automatic parameters were extracted, 75 were retained after dimensional reduction. Regression using the selected automatic parameters (number of sonants and occlusives recognized per second) provides a correlation of 0.83 between the predicted and actual score. This modeling is reliable (cross-validation: r=0.82). The correlation reaches 0.89 when including associated factors in the modeling, while maintaining a high reliability (r=0.70). Conclusions The use of automatic speech analysis allows a reliable prediction of the communication impairment. This study opens up new perspectives for the use and optimization of automatic speech recognition systems in clinical evaluation, and the consideration of functional needs expressed by the patients during their follow-up.
BALAGUER-thesis.pdf (1.2 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04221427 , version 1 (28-09-2023)

Identifiers

  • HAL Id : hal-04221427 , version 1

Cite

Mathieu Balaguer, Julien Pinquier, Jérôme Farinas, Virginie Woisard. Measurement of communication impairment after treatment for oral and oropharyngeal cancer by automatic analyses of spontaneous speech associated with biopsychosocial factors. 32nd World Congress of the IALP (2023), IALP: International Association of Communication Sciences and Disorders, Aug 2023, Auckland, New Zealand. ⟨hal-04221427⟩
23 View
2 Download

Share

Gmail Facebook X LinkedIn More