Date: June 2018
Source: Respirology, Volume 23, Issue 6, pp 560-561.
Abstract: Craniofacial structure is an important anatomical determinant of obstructive sleep apnoea (OSA). While obesity is the major attributing risk factor for OSA, it is well established that a restricted craniofacial bony skeleton also contributes to a more collapsible upper airway, which is a key pathogenic mechanism of OSA.1 Numerous studies have examined craniofacial and upper airway structures using imaging, including cephalometry and magnetic resonance imaging/computed tomography (MRI/CT).2, 3 These have led to much greater understanding of the specific anatomic risk factors for OSA, such as a smaller mandible/maxilla, cranial base abnormalities, inferior displacement of the hyoid and excess of certain upper airway soft tissues. However, imaging is often not performed as part of a clinical assessment due to its expense, radiation exposure and often requiring time‐consuming analysis for quantification. It is on this premise that facial photographic imaging could potentially be developed as a tool for craniofacial assessment and may be clinically useful in OSA prediction or risk stratification. Early work using standard frontal and profile facial photographs in a Caucasian population demonstrated the potential for the modality with OSA prediction accuracy to be around 76%.4 Similar prediction characteristics were also shown in a Chinese population, supporting its broader applications in other ethnic groups.5
In this issue of the journal, Lin et al.6 describe the application of a three‐dimensional (3D) photography imaging tool to evaluate facial profiles in patients with OSA. Subjects were recruited from a single centre in Taiwan. Facial images were captured on standard digital photographs, 3D photographs using a dedicated system for three‐dimensional stereo‐photogrammetry (3dMD) and spiral CT scan of the head and neck. Quantitative measurements of the various facial regions were performed. Cross comparisons of these measurements show much stronger agreement between the CT scan and 3D photographs (mean Cronbach’s alpha values >0.9), than with the standard two‐dimensional (2D) photographs. Correlation between Apnoea‐Hypopnoea Index (AHI) and some facial measurements were demonstrated with all three modalities of imaging. A multivariate model for OSA severity was presented for the 3dMD analysis showing modest predictive power (r = 0.52), but direct comparison of overall OSA predictive performance was not made with the other imaging methods. The authors concluded that 3D facial photography is highly accurate and would offer opportunities for applications in OSA.
There are notable advantages with 3D imaging of the face over standard photographs, including reduction in distortion and projection errors, ease of subject alignment and overall improved landmark identification. The accuracy and clinical application of the technology has been shown in other fields as highlighted in the publication.6 The fundamental question is what are the applications for such technology in the field of sleep medicine?
One proposed application for detailed craniofacial assessment is in the prediction of OSA. Despite data suggesting facial photographic prediction achieving at least a similar level of prediction to other OSA clinical prediction tools, it is unable to definitively rule in/out OSA.4, 5 Surface facial measurements do not differentiate between soft tissues and underlying skeletal structures, thus unable to truly assess anatomical imbalance to make inference on the degree of airway compromise.7 Even detailed imaging of the airway using gold standard techniques of MRI/CT does not have the predictive power for this application, as non‐anatomical pathophysiological factors are not considered.8, 9 Hence, 3D facial imaging alone would be unlikely to improve OSA prediction.
Another approach of applying facial photographic prediction is as a triage tool to identify patients at high/low risk of OSA. Similar to other prediction tools, this might have specific clinical utility to reduce undiagnosed disease burden or identify those with high pre‐test probability for home‐based testing.10 One such potential novel application could be in drivers’ licencing where facial photographs are routinely taken and may offer an opportunity for OSA risk stratification and screening. Machine learning techniques have been applied for automatic detection of facial photographs to accurately predict other conditions with distinct craniofacial features.11 Similar innovations may be translated to OSA prediction and further research is needed to determine the clinical applications for such technology.
Advances in understanding of the pathophysiology of OSA have led to efforts to develop clinically useful methods for disease phenotyping and explore new treatment paradigms.12 Another important application for facial photography is in craniofacial phenotyping to identify anatomical characteristics that are important in OSA pathogenesis. While physiological traits such as upper airway collapsibility have not been studied in relation to facial morphology, there is strong evidence that certain surface facial dimensions relate to upper airway structures (e.g. tongue size on MRI), which in turn influence airway dimensions and collapsibility.13, 14 Facial morphology has also been shown to be highly correlated with measures of central obesity.15 Facial measurements have been used as a method for stratification of the degree of anatomical compromise to investigate relationships to other OSA physiological traits.16
Collectively, these studies illustrate that facial photography could be an important modality for anatomical phenotyping. Given its availability, ease of use, high throughput and potential for analysis automation, these characteristics are ideal for applications in the epidemiological and longitudinal settings. As we move towards a personalized approach to diagnosis and treatment in OSA, developing a tool that can reliably quantify the craniofacial anatomy is likely to have immense clinical utility.
Article: Three‐dimensional facial phenotyping in obstructive sleep apnoea.
Authors: Richard WW Lee MBBS, PhD, FRACP.
Date: June 2018