Developing an Artificial Intelligence Tool to Predict Vocal Cord Pathology in Primary Care Settings
Bibliographic record
Abstract
OBJECTIVES: Diagnostic tools for voice disorders are lacking for primary care physicians. Artificial intelligence (AI) tools may add to the armamentarium for physicians, decreasing the time to diagnosis and limiting the burden of dysphonia. METHODS: Voice recordings of patients were collected from 2019 to 2021 using smartphones. The Saarbruecken dataset was included for comparison. Audio files were converted to mel-spectrograms using TensorFlow. Diagnostic categories were created to group pathology, including neurological and muscular disorders, inflammatory, mass lesions, and normal. The samples were further separated into sustained/a/and the rainbow passage. RESULTS: Two hundred three prospective samples and 1131 samples were used from the Saarbruecken database. The AI detected abnormal pathology with an F1-score of 98%. The artificial neural network (ANN) differentiated key pathologies, including unilateral paralysis, laryngitis, adductor spasmodic dysphonia (ADSD), mass lesions, and normal samples with 39%-87% F-1 scores. The Calgary database models had higher F-1 scores in a head-to-head comparison to the Saarbruecken and combined datasets (87% vs. 58% and 50%). The AI outperformed otolaryngologists using a standardized test set of recordings (83% compared to 55% ± 15%). CONCLUSION: An AI tool was created to differentiate pathology by individual or categorical diagnosis with high evaluation metrics. Prospective data should be collected in a controlled fashion to reduce intrinsic variability between recordings. Multi-center data collaborations are imperative to increase the prediction capability of AI tools for detecting vocal cord pathology. We provide proof-of-concept for an AI tool to assist primary care physicians in managing dysphonic patients. LEVEL OF EVIDENCE: 3 Laryngoscope, 133:1952-1960, 2023.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".