br on visual analog scale inteldesc
on visual analog scale; inteldesc: intelligibility on picture description; intelread: intelligibility on reading.
Fig. 2. Scatter diagram of results for signal impairment on picture description (sigdesc) versus signal impairment on reading (sigread): r = 0.89 (P < 0.001).
Please cite this article in press as: Balaguer M, et al. Assessment of impairment of intelligibility and of speech sig-nal after oral cavity and oropharynx cancer. European Annals of Otorhinolaryngology, Head and Neck diseases (2019), https://doi.org/10.1016/j.anorl.2019.05.012
M. Balaguer et al. / European Annals of Otorhinolaryngology, Head and Neck diseases xxx (2019) xxx–xxx 5
intelligibility than signal impairment severity on picture descrip-tion (0.77 vs. 0.69).
That mean scatter was wider for signal impairment than intel-ligibility, whether on discrete or continuous variables, argued for use of signal impairment assessment.
The picture description test involved semi-spontaneous speech, directed only by the picture and the instructions, and gave percep-tual scores with less ceiling effect for absence of disorder than the reading test. Mean scores were lower for semi-spontaneous speech, as seen in Fig. 1, and scatter was less.
The use of a graduated intelligibility scale, unlike in previous studies by the same team in a Cycloheximide with more varied pathol-ogy including neurologic dysarthria, has not been validated ; the ceiling effect was greater, with less scatter.
Thus, the present results as a whole confirmed the study hypoth-esis: assessing speech disorder severity from semi-spontaneous speech based on picture description is more clinically contributive than from reading.
Although the correlations between the various perceptual scores for head and neck cancer patients showed that their mean values all represent the same information, intelligibility showed a ceiling effect, increasing the scores by more than 1.5 points com-pared to signal impairment scores. Inter-observer reliability was better for intelligibility on the picture description task, but this was the only exception to the general superiority of measuring sig-nal impairment rather than intelligibility in description tests. The higher ICCs associated with intelligibility assessments may be due to the ceiling effect, with panelists tending more readily to give a maximum score of 10/10. Moreover, an ICC value of 0.69 for per-ceptual scores of speech signal impairment on picture description is enough for genital herpes parameter to be chosen .
Speech disorder is a frequent sequela in oral cavity and oropha-ryngeal cancer. Measurement is a challenge to be faced in seeking to improve management. The instruments used in everyday clinical practice are inherently subjective, depending on a listener’s percep-tion in assessing speech production deficit. The present results are in favor of using a measurement of signal impairment in a semi-spontaneous speech task, with severity defined as the degree of overall impairment of the vocal signal and the speech sample being based on a photograph randomly chosen from a set of 10 pictures all showing sea-fronts with one or several boats. The inter-observer reliability and metric properties of this parameter are compatible with clinical application in consultation for persons with experi-ence of this kind of disorder. Its limitations, however, argue for developing a disorder index based on automated speech analysis, to avoid intra- and inter-individual measurement variability.
Disclosure of interest Acknowledgments
The present study was part of the Carcinologic Speech Sever-ity Index (C2SI) project, funded by the Institut National du Cancer (funding no 2014-135) as a free research project in Human and Social Science, Epidemiology and Public Health. The principal inves-tigator is Virginie Woisard (Hôpital Larrey, Toulouse University Hospital).
We thank the medical and paramedical teams of the Institut Universitaire du Cancer Oncopole de Toulouse for recordings, and the listening panelists and patients who took part, and the Institut de Recherche en Informatique de Toulouse (Toulouse Information Technology Research Institute, Paul Sabatier Toulouse III Univer-sity) which was also involved.
 Mlynarek AM, Rieger JM, Harris JR, O’Connell DA, Al-Qahtani KH, Ansari K, et al. Methods of functional outcomes assessment following treatment of oral and oropharyngeal cancer: review of the literature. J Otolaryngol Head Neck Surg 2008;37(1):2–10.
 Lindblom B. On the communication process: speaker-listener interaction and the development of speech. Augment Altern Commun 1990;6(4):220–30.
 Hustad KC. The relationship between listener comprehension and intelligibil-ity scores for speakers with dysarthria. J Speech Lang Hear Res 2008;51(3): 562–73.