Evaluation of voice features using Wav2vec2.0-based deep learning as a diagnostic tool in acromegaly
Authors' conclusion
Voice analysis Wav2Vec2.0-based model demonstrates good diagnostic accuracy for acromegaly. These preliminary findings suggest that voice may serve as a promising non-invasive digital biomarker, pending validation in larger and more diverse cohorts.
Study context
- Population
- Patients with acromegaly and age- and sex-matched controlspatients with acromegaly 70 · age- and sex-matched controls 68
- Index test or model
- Wav2Vec2.0-based voice analysis against acromegaly status defined by standard biochemical criteria, symptoms, comorbidities, and residual tumor
- Target condition
- acromegaly
Methodological detailsage- and sex-matched controls, 80%/20% train-test split with five-fold cross-validation
- Age- and sex-matched controls
- 80%/20% train-test split with five-fold cross-validation
Performance
As reported in the abstract
Logisitic Regression model AUC
0.84
Logisitic Regression model specificity
0.92
Logisitic Regression model sensitivity
0.71
Logisitic Regression model Brier score
0.165
95% CI 0.057–0.289
Further findings
Decision curve analysis
- LR provided the highest benefit across a clinically relevant range of threshold probabilities
- ≈15-85%
Demographics
- Acromegaly mean age
- 45.5 years
- Acromegaly female
- 57.1%
- Control mean age
- 47.5 years
- Control female
- 61.8%
- Demographics were similar between groups
Original publication
Evaluation of voice features using Wav2vec2.0-based deep learning as a diagnostic tool in acromegaly. Pituitary. Published 13 November 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →