ABSTRACT
OBJECTIVES: To develop and validate machine learning (ML) and deep learning (DL) models using drug-induced sleep endoscopy (DISE) images to predict the therapeutic efficacy of hypoglossal nerve stimulator (HGNS) implantation. METHODS: Patients who underwent DISE and subsequent HGNS implantation at a tertiary care referral center were included. Six DL models and five ML algorithms were trained on images from the base of tongue (BOT) and velopharynx (VP) from patients classified as responders or non-responders as defined by Sher's criteria (50% reduction in apnea-hypopnea index (AHI) and AHI < 15 events/h). Precision, recall, F1 score, and overall accuracy were evaluated as measures of performance. RESULTS: In total, 25,040 images from 127 patients were included, of which 16,515 (69.3%) were from responders and 8,262 (30.7%) from non-responders. Models trained on the VP dataset had greater overall accuracy when compared to BOT alone and combined VP and BOT image sets, suggesting that VP images contain discriminative features for identifying therapeutic efficacy. The VCG-16 DL model had the best overall performance on the VP image set with high training accuracy (0.833), F1 score (0.78), and recall (0.883). Among ML models, the logistic regression model had the greatest accuracy (0.685) and F1 score (0.813). CONCLUSION: Deep neural networks have potential to predict HGNS therapeutic efficacy using images from DISE, facilitating better patient selection for implantation. Development of multi-institutional data and image sets will allow for development of generalizable predictive models. LEVEL OF EVIDENCE: N/A Laryngoscope, 2024.
ABSTRACT
The Activity-Based-Checks of Pain (ABCs) is a pain assessment tool incorporating activities of daily living and instrumental activities of daily living. Unlike widely used pain scales which are oftentimes unidimensional and highly subjective, the ABCs was designed to focus on function capabilities and limitations of patients due to pain. This study sought out to validate the factorial structure of the ABCs and assess its use in participants with chronic pain. Participants were recruited in two phases from Prolific - an online service designed to identify research participant recruitment based on study criteria. Phase one optimized the design of the ABCs, with 297 subjects selecting their preferred icon for each function and rating its understandability. The most preferred and understandable icons were then used in phase two, where 304 chronic pain participants completed the ABCs, PROMIS-29, additional PROMIS items that were analogous to the ABCs functions but not represented in the PROMIS-29, and the Brief Pain Inventory (BPI). Data was analyzed using exploratory factor analysis and confirmatory factor analysis demonstrating four factor loadings: multi-planal activities, sitting/hip flexor pain, walking/ambulation, and pain interference with lightweight unilateral activities. High internal consistency was demonstrated with all four factor loadings. Correlations between items in the ABCs, PROMIS, and BPI resulted in moderate to strong correlations demonstrating strong evidence for the validity of the ABCs as a functional pain assessment tool.
