Developing an Artificial Intelligence Tool to Predict Vocal Cord Pathology in Primary Care Settings.

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.

Effect of socioeconomic status on wait times for patients undergoing treatment for laryngeal conditions in a universal healthcare system.

Objectives: As one of the world's only fully publicly administered, universal healthcare systems, Canada intends to provide equitable access to services for all patients. Socioeconomic status (SES) can affect treatment wait times with implications on health outcomes; however, this has not been evaluated in the setting of laryngeal disease, which incorporates urgent and elective conditions, in a universal healthcare system. This study aims to identify whether SES-affected treatment wait times for laryngeal therapies in this system. Methods: A retrospective review was conducted on a cohort of patients with laryngeal disease at an academic, tertiary center who underwent laryngeal surgery over a three-year period. Data retrieved through medical records; surgical and voice therapy wait times were normalized to each practitioner's average wait time for respective diagnostic categories. Income was used to assess SES and was derived from Statistics Canada census information based on patient postal codes. Results: Data analysis identified 578 patients (59% male). Mean wait time to surgery for all conditions was 123.5 (95% confidence interval 113.1-133.9) days. Analysis of variance analysis found no difference in wait times between different SES groups (p = .28), regardless of laryngeal disease category. Patients with cancer or airway obstruction had shorter wait times compared with benign conditions (p < .0001). Conclusions: SES did not affect treatment wait times for laryngeal therapies in a universal healthcare system. Wait times were shorter for oncologic and obstructive conditions compared with benign conditions, reflecting an ability to accommodate clinical needs without impacting care access at the detriment of different SES statuses. Level of Evidence: 4.

Long-Term Voice Outcomes Following Cricotracheal Resection for Subglottic Stenosis: A Retrospective Analysis.

OBJECTIVES/HYPOTHESIS: Cricotracheal resection (CTR) is an effective treatment for refractory idiopathic subglottic stenosis (iSGS) but is associated with persistent dysphonia. Outcomes were measured longitudinally to characterize how patients' voices and self-perceived voice handicaps changed after CTR. STUDY DESIGN: Retrospective case-series. METHODS: We conducted a retrospective cohort study of patients with refractory iSGS treated by CTR from 2006 to 2017. Voice Handicap Index (VHI), Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V), and acoustic analysis were prospectively collected preoperatively and postoperatively at 1, 3, 6, 12, and 24-month intervals. A linear mixed model was used to evaluate temporal change. RESULTS: Thirty-three patients (97% female) were included. VHI scores increased significantly from baseline preoperative score (27.2 ± 22.7) to a mean value of 44.3 ± 25.6 (P < .001) 1-month postoperatively but decreased below preoperative scores after 2 years (18.8 ± 11.9, P = .795). Mean fundamental frequency (F0) values in speech decreased significantly from 192.0 ± 24.9 Hz preoperatively to 167.1 ± 19.8 Hz at a 2-year follow-up (P = .002), with a nadir value at 1-month postoperatively (148.4 ± 20.5, P < .001). CAPE-V scores increased significantly from preoperative to 1-month postoperative (3.0 ± 2.3 vs. 21.9 ± 17.3, P < .001) but returned toward baseline values at 24 months after CTR (6.9 ± 4.8, P = .027). CAPE-V measurements postoperatively were correlated with VHI and F0 (Pearson coefficient = 0.54 (VHI), -0.46 (F0), P < .001). CONCLUSIONS: Following CTR, mean F0 values were significantly and consistently lower but did increase over time, correlating with patients' improving VHI and CAPE-V scores. VHI values indicate that patient's perception of their voice is not significantly impacted in long term. These results provide a framework to counsel patients about long term voice expectations. LEVEL OF EVIDENCE: 4 Laryngoscope, 132:1054-1060, 2022.

Use of a Low-fidelity simulator to improve trans-nasal fibre-optic flexible laryngoscopy in the clinical setting: a randomized, single-blinded, prospective study.

BACKGROUND: Trans-nasal flexible fibre-optic laryngoscopy (TFFL) is an essential skill for otolaryngologists. There is evidence to suggest that simulators help residents acquire procedural skills. The objective of this study was to examine the effect of simulation on endoscopy skill acquistion. METHODS: A randomized controlled trial was conducted utilizing medical students and junior residents with limited experience in TFFL. Learners all performed a baseline endoscopy and were then randomized to receive either 45 minutes of simulation training or not. Following this, a second endoscopy was performed. Time to adequate visualization of the glottis, the percentage of time adequate visualization of the airway was maintained, and the number of collisions with mucosa were analyzed. Qualitative assessments were also obtained from the learner, patient, and staff laryngologist. RESULTS: Time to adequate visualization of the glottis and the number of mucosal collisions were significantly less during the second endoscopy, irrespective of the use of simulation (84.8 sec vs. 68 sec, p < 0.01; 5.0 vs. 3.2, p < 0.01, respectively). Analysis using a two-way ANOVA with interaction established that none of the quantitative measures analyzed in this study improved with the addition of simulation. CONCLUSION: Improvements in time to visualization of the glottis and number of mucosal contacts were seen between the first and second endoscopy irrespective of simulator use. No additional benefit was conferred with the use of a low-fidelity simulator.