Head position classification of medical imaging studies: an assessment and development of a protocol.

ABSTRACT

OBJECTIVES: To determine the optimal approach to reliably classify head position of head and neck medical imaging studies as flexion, neutral or extension for use in craniofacial and orthodontic research. METHODS AND MATERIAL: A prospective study scanned six participants in flexed, neutral and extended head positions. Additionally, a retrospective dataset of 46 CT studies were visually classified into six categories flexion, neutral-flexion, neutral, neutral-extension, extension and flexion-extension. 14 landmarks were placed in the head and neck region of all studies to calculate 17 head position angle and distance measurements. Assessment of head position classification was performed for each measure, as well as all measures together using GUIDE forest. RESULTS: No single measure was sufficient to reliably classify head position in both retrospective and prospective imaging studies. Therefore, this study developed a head position protocol that considers multiple measures using two hybrid predictive models, to classify head position. Compared to visual assessment of head position, this protocol classified the imaging studies into the four head position categories with 82% neutral sensitivity and 100% neutral precision where the three neutral groups (neutral-flexion, neutral and neutral-extension) were grouped together. CONCLUSION: This study established a novel head position classification protocol that uses multiple measures accounting for both head and neck positions to reliably classify head positions in imaging studies as flexion, neutral or extension. Given the limitation that no single measure reliably classified head position, this protocol is strongly recommended to researchers who need to account for head position to reach valid conclusions.