Jpetiot/ January 8, 2014/ Previous

Carcinologic Speech Severity Index Project

Main issues and objectives

C2SI: Carcinologic Speech Severity Index. A Speech disorders Severity Index to measure the impact of oral and pharyngeal cavity on speech production.


The decreasing mortality of Head and Neck cancers (HNC) highlights the importance to reduce the impact on quality of life (QoL). Nevertheless, the usual tools for assessing QoL are not relevant for measuring the impact of the treatment on the main functions involved by the sequellae. Validated tools for measuring the functional outcomes of carcinologic treatment are missing, in particular for speech disorders. Some assessments are available for voice disorders in laryngeal cancer but there are based on very poor tools for oral and pharyngeal cancers involving more speech articulation issues than voice issues.

In this context, we propose to develop a severity index of speech disorders to describe the impact of therapeutic protocols with the aim to complete the survival rates.

Speech Intelligibility is the usual way to quantify the severity of neurologic speech disorders.  But this measure is not valid in clinical practice because of several difficulties as the “familiarity effect” and poor inter-judge reproducibility. Moreover, the transcription intelligibility scores do not accurately reflect listener comprehension.

Recently, it is acknowledged that an unbiased and objective assessment of the communication deficiency caused by a speech disorder calls for automatic speech processing tools. The idea is to collect audio recordings of patient speech and to compute an intelligibility score of the utterances. Middag et al in 2012 presented a new method that predicts running speech intelligibility in a way that is robust against changes in the text and against differences in the accent of the Dutch speakers applicable to patients treated for HNC.


Therefore, our hypothesis is that an automatic assessment technique can measure the impact of speech disorders on the communication abilities giving a severity index in speech production of patients treated for head and neck and particularly for oral and pharyngeal cancer. We will name this index the Carcinologic Speech Severity Index (C2SI).

Research work’s main objective

The main objective is to demonstrate that the carcinologic speech severity index (C2SI), obtained by an automatic speech processing tool, produces equivalent or superior outcomes than a speech intelligibility score obtained by human listeners, in terms of quality of life predicing the speech handicap, after treatment of oral and/or pharyngeal cancer.

Secondary objectives:

  1. To compare this predicitive model to  intelligibility and comprehension scores taken as references
  2. To describe the profile of a patient in terms of speech characteristics, anatomic and functional impairment
  3. To contribute to a specific speech database of carcinologic patients (parolothèque), recording minimal interviews in the aim to do future research in several domains such as linguistic, psychology and/or epidemiology.


The design of this study is transversal, without restriction for participating in other studies. The methodology follows language sciences rules adapted to oncology. We expect as described par Middag 2012, a correlation at 0,8 +/- 0,15 between the automatic score and the intelligibility score obtained by the jury. 94 patients would be necessary on this criterion. Considering a risk of 10% for having a matter with a sound file, the recruitment will be of 103 within three Toulouse departments (ENT, Cancerology and Maxillofacial).

The inclusion criteria are for the population of the study :

  • Oral and Pharyngeal cancer T2 to T4 for having a wide dynamic of deficit
  • Treatment by surgery and/or radiotherapy and/or chemotherapy
  • Delay after the end of the treatment of minimum 6 months. No maximal limit because the time of inclusion must be in link with the survival.

The main Exclusion criteria are: patients with a previous speech disorders (ie stuttering… )and patients not able to perform the whole task. A control group of 35 normal speakers is necessary for having a training group for the automatic assessment technic.

Five steps will be necessary:

  1. Corpus constitution by (1) recording the minimal interview, a standardized text and lists of words read by the patients, (2) gathering others criteria : Speech Handicap Index scores (SHI), motor impairment scores obtained by a logopedic assessment tool, generic Quality of Life scale (EORTC C30 & HN35).
  2. Preparation of the stimuli extracted from the recordings and perceived intelligibility and comprehension tests.
  3. Preparation of the final database for the computer science teams with the perceived intelligibility and comprehension scores, the SHI score and other QoL measures, individual data, carcinologic data…
  4. Determination of the predictive modeles of intelligibility, of comprehension and others combination leading to the choice of the C2SI regarding the correlations with the Speech Handicap

  5. Description of the patient profile regarding the risk level of speech handicap, in the aim for proposing the validation of the C2SI using this parameter for decribing the outcomes in therapeutic trials and after in practice.

Expected results

The delivrables will be predictive models giving  index of intelligbility,  comprehensibility, and severity. These index can be the object of patents. The results will also include the description of  profiles of speech disorders and patients with medical and technological perspectives of development

Relevance and originality of the project regarding the state of the art:

The originality of this project is to open a research on tools assessing functional results between QoL and anatomic modifications induced by the cancer and its treatments. This kind of validated tools could complete survival rate for describing and comparing the results of therapeutic protocol in oncology for tumors leading to a high level of handicap. Using an automatic speech processing tool combining linguistic and computational sciences applicable to clinical practice in oncology will give a high level of reliability and objectivity

Working on speech disorders is a real challenge because of the collaboration with linguistic laboratories and the necessity to deal with several languages. Another originality is to use an approach allowing European collaborations and the possibility of freeing itself from linguistic “barriers”.


People involved in the SAMOVA team



  • Start time : 1st november 2014
  • End time : 31th december 2018
Share this Post