"The Relationship Between the Cepstral/Spectral Index of Dysphonia (CSID) and CAPE-V Ratings : A study using Japanese Speech Tasks"

【Background】Acoustic analysis, alongside auditory-perceptual evaluation, plays a critical role in the assessment of voice disorders. The Cepstral/Spectral Index of Dysphonia (CSID) has been widely recognized for its utility, showing a strong correlation with the auditory perceptual evaluation method, CAPE-V. However, the optimal weighting of CSID values derived from vowels and sentences (easy onset, all voiced, hard-glottal attack, and voiceless plosive; referred to as CSID_v, CSID_e, CSID_av, CSID_h, and CSID_vp, respectively) for evaluating the degree of voice disorders remains unclear defined.

【Objective】This study aimed to investigate the influence of CSID values on CAPE-V measures—overall severity, roughness, and breathiness (CAPE-V_os, CAPE-V_r, and CAPE-V_b, respectively)—using multiple regression analysis. The dependent variables were CAPE-V measures, and the independent variables were CSID values from sustained vowels and sentences.

【Methods】125 adults (95 with voice disorders and 30 without voice problems) were instructed to produce sustained vowels and read Japanese sentences. Recordings were made using a headset microphone, and CSID were calculated using ADSV (Analysis of Dysphonia in Speech and Voice) software integrated into the CSL Model 4300 (PENTAX Medical,Montvale,NJ). CAPE-V assessments were performed by three clinicians specializing in voice disorders, with the average of their scores used for analysis. Using forced-entry multiple regression analysis, variables with p-values > 0.05 were excluded, and further regression analyses were conducted.

【Results】For CAPE-V_os and CAPE-V_r, only CSID_v and CSID_e were significant predictors. The relative contributions were CSID_e > CSID_v for CAPE-V_os and CSID_e < CSID_v for CAPE-V_r. The adjusted R²values were 0.666 (R=0.816) for CAPE-V_os and 0.506 (R=0.771) for CAPE-V_r, indicating strong model fit. The regression equations were:

CAPE-V_o=0.668×(CSID_e)+0.377× (CSID_v)+18.201

CAPE-V_r =0.347×(CSID_e)+0.334×(CSID_v)+10.387　

For CAPE-V_b, only CSID_e was a significant predictor(R=0.771, adj.R²=0.595), with the equation: CAPE-V_b=0.933×(CSID_e)+16.886

【Conclusion】CSID_v and CSID_e were identified as significant variables influencing Japanese CAPE-V_os, CAPE-V_r, and CAPE-V_b scores, with CSID_e having the greatest impact across all measures. These findings suggest the importance of CSID_e in the analysis of Japanese speech. However, because the study included many cases of nodules, polyps, or vocal fold paralysis, future research will need to investigate whether other CSID values contribute to conditions such as spasmodic dysphonia or functional voice disorders.