Development and validation of a computer-assisted stuttering analysis framework for objective severity assessment

Abstract Background Standardized assessment of stuttering severity remains challenging due to reliance on perceptual judgment and manual scoring. Computer-assisted analysis may improve consistency and efficiency by providing objective quantification of stuttering parameters. Objective To develop and validate the Computer-Based Stuttering Analysis (CBSA) framework for computer-assisted scoring of the Arabic Stuttering Severity Instrument (ASSI), with additional fluency-related measures. Methods This cross-sectional study included 160 Arabic-speaking participants: 80 individuals with developmental stuttering (40 readers and 40 non-readers) and 80 matched fluent controls. Speech samples were collected following ASSI procedures and manually annotated using Praat. CBSA processed the annotations to compute ASSI-based measures of stuttering frequency, Mean Duration of the Three Longest Stuttering Segments (MDTLSS), and severity scores, along with fluency parameters including Speech Rate (SR), Speech Efficiency Score (SES), and Total Speech Time (TST). Validity, agreement, reliability, and diagnostic performance were evaluated using correlation analysis, Bland–Altman plots, ROC analysis, and reliability statistics. Results CBSA showed strong agreement with ASSI measures of frequency, duration, and severity (Spearman’s r = 0.87–0.98, P < 0.01), with minimal bias on Bland–Altman analysis. Fluency measures significantly differentiated patients from controls ( P < 0.01) and showed high diagnostic accuracy (AUC = 0.95 for SR, 0.96 for SES, and 0.95 for TST). Internal consistency was acceptable (Cronbach’s α = 0.68–0.75), while inter-rater reliability was high (ICC = 0.93–0.97). Conclusion CBSA is a reliable computer-assisted framework for objective assessment of stuttering severity and fluency in Arabic-speaking populations.