ABSTRACT Airway abnormalities in children, such as subglottic stenosis (SGS) and Robin sequence (PRS), may result in breathing difficulties, risk of recurrent infections, hypoxia, respiratory insufficiency, life-threatening events, and long-term morbidity. In children with airway abnormalities, a multidisciplinary approach to care involves selection from a variety of medical and surgical interventions. Therapy is typically directed by the clinician's experience and preference, rather than based on normalized, quantitative physiologic and anatomic metrics. Static computed tomography (CT), dynamic CT, and bronchoscopy have been considered for quantitative diagnosis and assessment. However, quantitative measures of what constitutes normal airway geometry and how normal airway geometry changes with respect to age, weight, and sex are lacking. Such normative measures can be used to score the degree of airway abnormality, define thresholds for abnormality, and better understand surgical interventions' impact. In previous work, we developed the Pediatric Airway Atlas to provide spatially localized normative measures for upper airway cross-sectional areas in children derived from a population of static 3D CT images. The goal of the proposed study is to build upon our database of 3D CT images and associated clinical measures to develop the computational methodology for a Pediatric Airway Shape Atlas (PASA), which will model the upper airway as a 3D shape instead of restricting airway characterization to cross-sectional area only. The PASA will allow for a comprehensive characterization of 3D geometry. Specifically, the core of the PASA will be a new, innovative neural additive shape model that is designed to allow for interpretable results, captures the effects of relevant covariates (such as age, sex, and weight), and allows within the same framework to predict likely airway changes over time for individuals thereby providing a means to quantify the effect of surgical interventions on 3D airway geometry. Our approach will provide improved, non-invasive quantification of airway abnormalities. Automated data analysis will allow for rapid refinement of atlas-based analyses and will greatly simplify use by other research and clinical groups. The resulting software will be open-source. Furthermore, the new methodologies developed will be broadly applicable to multiple, common causes of airway obstruction in children and adults.