Projet DAPADAF-E

Validity of a task of acoustic-phonetic decoding on anatomic deficits in paramedical assessment of speech disorders for patients treated for oral or oropharyngeal cancer

Cancers of the upper aerodigestive tract is the fifth most common cancer in terms of incidence in France (HAS 2009). Due to their location, they impact the patients’ speech abilities (Mlynarek, 2008).

In care management, speech-language pathology clinical assessments include two components. The analytic part allows identifying motor or sensitivity disorders of the organs involved in speech production. The functional part characterizes the impact on speech and communication using an assessment “at the ear.” Even if in oncology the link between anatomy and functional deficits in speech is very tight (the location of the tumor, its size and the structural modifications due to treatment modify the oro-pharyngeal dynamics and thus the mechanisms involved in speech production), the correlation between intelligibility scores and motor scores is low (Lazarus, 2013). In other words, there is a weak link between motor deficits and their consequences on speech production, mainly due to speaker compensation (the deficits are circumvented by the production of phonemes close to the target phoneme) or the interlocutor. Indeed, the perception of speech by a human listener is not a simple “recording” of the production. It is more like a representation of this production after the implementation of individual mechanisms for restoring acoustic information related to the lexicon or context by the listener (Ghio, 2019). The listener’s degree of familiarity with the speaker or his/her pathology is also a source of variability (Fex, 1992).

The gap between speech production and perception by the interlocutor could be managed by the use of a more adapted and automatic task. An acoustic-phonetic decoding task (production of isolated pseudowords in repetition or reading), created within the framework of the InCA C2SI project, avoids the effects of restoration. An automatic score derived from the acoustic-decoding task would result in an overall score per patient, but also in specific scores for each phonetic segment, to be correlated with the anatomical analytical segment scores.

Thus, the hypothesis is that the automatic processing of an acoustic-phonetic decoding task during the assessment in current practice is a tool for the diagnosis of analytical and dynamic oropharyngeal deficits by highlighting deficient linguistic units. In other words, an automatic treatment of speech in oncology on a pseudoword production task would allow a more efficient link between disturbances in speech production and motor impairment of speech organs.

Liste des partenaires :

  • Hospital of Toulouse/Voice and Swallowing Unit, Intensive Rehabilitation Unit for Laryngectomized Patients: Mathieu Balaguer (scientific coordinator), Clémence Devoucoux, Anaïs Galtier, Virginie Woisard
  • Assistance Publique Hôpitaux de Marseille, Hospitals of Marseilles : Nicolas Fakhry, Camille Galant, Danièle Robert, Alexia Mattei
  • IUCT Oncopole : Pauline Chaussier
  • Laboratoire Parole et Langage (Aix-en-Provence) : Alain Ghio, Muriel Lalain, Marie Rebourg
  • Laboratoire Informatique d’Avignon : Sondes Abderrazek, Corinne Fredouille

Coordinateur du projet : Mathieu Balaguer (IRIT)

Responsable Scientifique pour le LIA : Corinne Fredouille

Période : 2020-2024

En savoir plus