Introduction:
ENT clinics frequently see cases of vocal disorders. Voice abnormalities in the Saudi Arabian setting have, however, been the subject of little investigation. This retrospective study aimed to analyse vocal recordings from Al-Moosa Hospital in Al-Hasa, Saudi Arabia, to distinguish between healthy and pathological voices.
Methods:
A study involving 208 adult patients was conducted to analyse sustained vowel phonation using an iPhone 14 Pro Max. Participants were categorised into eight groups based on diagnosis, and twelve acoustic parameters were extracted from the recordings. Pitch analysis was conducted to map F0 values to musical pitch notes and frequencies. Python libraries such as librosa and numpy were used to extract pitch data. Statistical analysis was performed using SPSS, with normality assessed using the Shapiro–Wilk test. Analyses included one-way analysis of variance, paired sample t-tests, correlations between variables, multiple linear regression and logistic regression. Statistical significance was set at P < 0.05.
Results:
Significant variations were seen in eight acoustic measures across all diagnosis categories. Pitch analysis characterised characteristics across groups by identifying average F0 ranges and mapping them to musical notes. Pitch mapping improved the characterisation of profiles. Polyps have a wider range of low pitches than nodules that maintain higher notes. Using statistical modelling, pathology prediction and determinants were connected, with noise-to-harmonics ratio demonstrating superior classification.
Conclusion:
The study identifies vocal fold pathologies in Saudi patients using objective acoustic analyses. It reveals significant variations between healthy and diseased voices, and a new pitch mapping enhancement makes profiles easier to distinguish. Statistical modelling consistently predicts diagnoses based on acoustic anomalies. Despite limitations, the study improves clinical knowledge and lays the groundwork for future research and real-world applications.