Artificial intelligence-driven computed tomography analysis of upper airway morphology in patients undergoing multilevel sleep surgery
Year: 2026
Type: article
Abstract: Anatomical information for planning sleep surgery is limited. We conducted an exploratory analysis to quantify upper airway morphology before and after multilevel sleep surgery and to examine whether preoperative measurements were associated with surgical response. In this post hoc analysis of a prospective cohort, 25 adults with moderate-to-severe obstructive sleep apnea underwent expansion sphincter pharyngoplasty, tonsillectomy, and single-port transoral robotic tongue base resection, with nasal surgery when indicated. Preoperative and postoperative computed tomography scans were segmented using a previously developed deep-learning pipeline. Surgical response was defined a priori as a postoperative apnea–hypopnea index below 15 events/h. Eighteen patients (72.0%) met the definition of response. Segmentation agreed with manual ground truth (mean Dice coefficient 0.868 ± 0.034). Surgery was associated with significant change in 8 of 13 morphological parameters, including uvula area (+ 140.2 mm²; 95% CI 109.2–171.3) and uvula–tongue volume (+ 2270.5 mm³; 95% CI 1166.6–3374.3). No preoperative parameter distinguished responders from non-responders (all p > 0.05), and preoperative tongue–epiglottis volume showed no discrimination. No drug-induced sleep endoscopy finding differed by response status. Automated computed tomography analysis reliably quantified upper airway remodeling after multilevel sleep surgery, but no preoperative morphological parameter predicted surgical response. These exploratory findings require independent confirmation.
Source: Scientific Reports
Language: English
Subfield: Physiology (medical)
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
Award: G8628864617
OpenAlex ID: W7225402017