Machine learning and geometric morphometrics to predict obstructive sleep apnea from 3D craniofacial scans

Evolving trends in novel sleep tracking and sleep testing technology publications between 2020 and 2022

STUDY OBJECTIVES

To update sleep medicine providers regarding (1) published research on the uses and performance of novel sleep tracking and testing technologies, (2) the use of artificial intelligence to acquire and process sleep data, and (3) research trends and gaps regarding the development and/or evaluation of these technologies.

METHODS

Medline and Embase electronic databases were searched for studies utilizing screening and diagnostic sleep technologies, published between 2020 and 2022 in journals focusing on human sleep. Studies' quality was determined based on the Study Design criteria of The Oxford Center for Evidence-Based Medicine Levels of Evidence.

RESULTS

Ninety-six of 3,849 articles were included. Most studies were adult performance evaluation (validation) studies, often comparing a novel technology to polysomnography. Sleep tracker publications tended to be Unites States-based, nonindustry-funded, performance studies on healthy adults using non-Food and Drug Administration-cleared technologies. Sleep apnea testing technologies were more frequently industry-funded and Food and Drug Administration-cleared. All studied technologies utilized software with an algorithm and/or artificial intelligence. Few studies used randomized control designs, or accounted for recruitment/attrition biases associated with participants' age, race/ethnicity, or comorbid health conditions.

CONCLUSIONS

Evidence-based publications have not kept pace with the proliferation and landscape of consumer and clinical sleep technologies. Due to the variance in technologies used within sleep research, careful review of the software used within studies is recommended. Future publications may fill identified gaps by including underrepresented populations, maintaining independence from industry, and through rigorous study design.

CITATION

Holfinger S, Schutte-Rodin S, Ratnasoma D, et al. Evolving trends in novel sleep tracking and sleep testing technology publications between 2020 and 2022. J Clin Sleep Med. 2025;21(5):891-905.

AI-Driven Detection of Obstructive Sleep Apnea Using Dual-Branch CNN and Machine Learning Models

Background/Objectives: The purpose of this research is to compare and contrast the application of machine learning and deep learning methodologies such as a dual-branch convolutional neural network (CNN) model for detecting obstructive sleep apnea (OSA) from electrocardiogram (ECG) data. Methods: This approach solves the limitations of conventional polysomnography (PSG) and presents a non-invasive method for detecting OSA in its early stages with the help of AI. Results: The research shows that both CNN and dual-branch CNN models can identify OSA from ECG signals. The CNN model achieves validation and test accuracy of about 93% and 94%, respectively, whereas the dual-branch CNN model achieves 93% validation and 94% test accuracy. Furthermore, the dual-branch CNN obtains a ROC AUC score of 0.99, meaning that it is better at distinguishing between apnea and non-apnea cases. Conclusions: The results show that CNN models, especially the dual-branch CNN, are effective in apnea classification and better than traditional methods. In addition, our proposed model has the potential to be used as a reliable, non-invasive method for accurate OSA detection that is even better than the current state-of-the-art advanced methods.

Upper airway and hyoid bone-related morphological parameters associated with the apnea-hypopnea index and lowest nocturnal oxygen saturation: a cephalometric analysis

BACKGROUND

Obstructive sleep apnea (OSA) is a common disorder characterized by repetitive complete or partial closure of the upper airway during sleep, resulting in sleep fragmentation and oxygen desaturation. Cephalogram is recognized as an effective diagnostic tool for predicting OSA risk in clinical practice. This study aims to assess and analyze the morphological characteristics of the upper airway and hyoid bone position associated with OSA using data from polysomnography studies and two-dimensional cephalometric analysis.

METHODS

The study included lateral cephalograms and polysomnography reports from the records of 105 adult (64 males & 41 females) patients who underwent comprehensive clinical examination. The severity of OSA was evaluated based on the apnea-hypopnea index (AHI) and lowest nocturnal oxygen saturation (LSaO2). The participants were divided into male and female groups to investigate the correlation between cephalometric parameters and OSA severity. Thirteen cephalometric parameters, including eleven linear measurements and two angular measurements, were analyzed. The significance level was set at P-value < 0.05.

RESULTS

The male group exhibited significantly higher severity of OSA compared to the female group, as indicated by higher AHI and lower LSaO2. There was an inverse association between AHI values with width of upper airway as well as distance between hyoid bone position relative to mandibular plane in both male and female groups. Additionally, only the male group showed a correlation between hyoid bone position relative to gonion/third-fourth vertebrae positions with AHI values. 4 out of 7 parameters associated with AHI in male group remained correlated with LSaO2, while in females only the distance between hyoid bone and line formed by ptergoid and pterygomaxillary fissure point showed correlation with LSaO2.

CONCLUSION

Correlation analysis revealed that a narrower upper airway was positively associated with increased AHI, while an inferiorly positioned hyoid bone in relation to mandible was negatively correlated with LSaO2. Our findings highlight the importance of several cephalometric parameters in predicting OSA severity based on AHI and LSaO2 levels; moreover, certain parameters exhibited significant gender-specific associations.

Artificial Intelligence Applications in Pediatric Craniofacial Surgery

Artificial intelligence is rapidly transforming pediatric craniofacial surgery by enhancing diagnostic accuracy, improving surgical precision, and optimizing postoperative care. Machine learning and deep learning models are increasingly used to analyze complex craniofacial imaging, enabling early detection of congenital anomalies such as craniosynostosis, and cleft lip and palate. AI-driven algorithms assist in preoperative planning by identifying anatomical abnormalities, predicting surgical outcomes, and guiding personalized treatment strategies. In cleft lip and palate care, AI enhances prenatal detection, severity classification, and the design of custom therapeutic devices, while also refining speech evaluation. For craniosynostosis, AI supports automated morphology classification, severity scoring, and the assessment of surgical indications, thereby promoting diagnostic consistency and predictive outcome modeling. In orthognathic surgery, AI-driven analyses, including skeletal maturity evaluation and cephalometric assessment, inform optimal timing and diagnosis. Furthermore, in cases of craniofacial microsomia and microtia, AI improves phenotypic classification and surgical planning through precise intraoperative navigation. These advancements underscore AI's transformative role in diagnostic accuracy, and clinical decision-making, highlighting its potential to significantly enhance evidence-based pediatric craniofacial care.

Shape analysis of the craniofacial skeleton in children prenatally exposed to anticonvulsant medications using geometric morphometrics

Children exposed prenatally to antiepileptic drugs may have a typical facies characterized by midfacial retrusion, a short nose, and anteverted nares. Our aim was to determine whether the shape of the maxilla was altered in its sagittal displacement, or whether the defect in the underlying articulation with the cranial base was responsible for the appearance of midface retrusion. Our hypothesis was that the sphenoid bone as well as the maxilla and other bones in the cranial base were affected by the anticonvulsant medication. The lateral cephalograms of 65 children exposed prenatally to monotherapy (phenobarbital, phenytoin, or carbamazepine) were evaluated using various analyses derived from geometric morphometrics (GM) on different studied areas (maxilla, entire cranial base, spheno-occipital region, and the total study area) and the resulting configurations compared with those of control children. Procrustes ANOVA suggested that shape variation for all the regions correlated significantly (p < 0.0001) with exposure to antiepileptic drugs, and principal component analysis revealed a noticeable separation between the means of the two groups when PC1 was plotted against PC2 for all the areas studied. The cross-validation resulting from the discriminant function analysis accurately classified between 79.5% and 88.6% of the control group and between 73.8% and 90.7% of the study group when looking at the different anatomic regions. Canonical variate analysis, applied to the sample after its separation following biological sex and stratification into two age groups, showed unequal results between males and females as well as during circumpubertal growth of the cranial base. Thus, in the exposed subjects, while the glabella was projected forward with a similar prominence in males and females, the rhinion, which is relocated more posteriorly, was more severely displaced in females as opposed to the sella, where the most important displacement occurred in males. Regarding the age groups, it revealed that patients in the younger group of both sexes exhibited a facial shape difference very early (p < 0.0001) when the comparison was performed between exposed and non-exposed subjects. This difference was maintained in females at older ages but not in males. These details may help isolate the mechanism for the anomalies because of GM's use of shape instead of traditional linear and angular cephalometric measurements.

Advancements in Obstructive Sleep Apnea Diagnosis and Screening Through Artificial Intelligence: A Systematic Review

BACKGROUND

Obstructive sleep apnea (OSA) is a prevalent yet underdiagnosed condition associated with a major healthcare burden. Current diagnostic tools, such as full-night polysomnography (PSG), pose a limited accessibility to diagnosis due to their elevated costs. Recent advances in Artificial Intelligence (AI), including Machine Learning (ML) and deep learning (DL) algorithms, offer novel potential tools for an accurate OSA screening and diagnosis. This systematic review evaluates articles employing AI-powered models for OSA screening and diagnosis in the last decade.

METHODS

A comprehensive electronic search was performed on PubMed/MEDLINE, Google Scholar, and SCOPUS databases. The included studies were original articles written in English, reporting the use of ML algorithms to diagnose and predict OSA in suspected patients. The last search was performed in June 2024. This systematic review is registered in PROSPERO (Registration ID: CRD42024563059).

RESULTS

Sixty-five articles, involving data from 109,046 patients, met the inclusion criteria. Due to the heterogeneity of the algorithms, outcomes were analyzed into six sections (anthropometric indexes, imaging, electrocardiographic signals, respiratory signals, and oximetry and miscellaneous signals). AI algorithms demonstrated significant improvements in OSA detection, with accuracy, sensitivity, and specificity often exceeding traditional tools. In particular, anthropometric indexes were most widely used, especially in logistic regression-powered algorithms.

CONCLUSIONS

The application of AI algorithms to OSA diagnosis and screening has great potential to improve patient outcomes, increase early detection, and lessen the load on healthcare systems. However, rigorous validation and standardization efforts must be made to standardize datasets.

A Survey Study of the 3D Facial Landmark Detection Techniques Used as a Screening Tool for Diagnosis of the Obstructive Sleep Apnea Syndrome

Obstructive Sleep Apnea (OSA) is a common disorder affecting both adults and children. It is characterized by repeated episodes of apnea (stopped breathing) and hypopnea (reduced breathing), which result in intermittent hypoxia. We recognize pediatric and adult OSA, and this paper focuses on pediatric OSA. While adults often suffer from daytime sleepiness, children are more likely to develop behavioral abnormalities. Early diagnosis and treatment are important to prevent negative effects on children's development. Without the treatment, children may be at increased risk of developing high blood pressure or other heart problems. The gold standard for OSA diagnosis is the polysomnography (sleep study) PSG performed at a sleep center. Not only is it an expensive procedure, but it can also be very stressful, especially for children. Patients have to stay at the sleep center during the night. Therefore, screening tools are very important. Multiple studies have shown that OSA screening tools can be based on facial anatomical landmarks. Anatomical landmarks are landmarks located at specific anatomical locations. For the purpose of the screening tool, a specific list of anatomical locations needs to be identified. We are presenting a survey study of the automatic identification of these landmarks on 3D scans of the patient's head. We are considering and comparing both knowledge-based and AI-based identification techniques, with a focus on the development of the automatic OSA screening tool.

Obstructive sleep apnea detection during wakefulness: a comprehensive methodological review

Obstructive sleep apnea (OSA) is a chronic condition affecting up to 1 billion people, globally. Despite this spread, OSA is still thought to be underdiagnosed. Lack of diagnosis is largely attributed to the high cost, resource-intensive, and time-consuming nature of existing diagnostic technologies during sleep. As individuals with OSA do not show many symptoms other than daytime sleepiness, predicting OSA while the individual is awake (wakefulness) is quite challenging. However, research especially in the last decade has shown promising results for quick and accurate methodologies to predict OSA during wakefulness. Furthermore, advances in machine learning algorithms offer new ways to analyze the measured data with more precision. With a widening research outlook, the present review compares methodologies for OSA screening during wakefulness, and recommendations are made for avenues of future research and study designs.

Achieving Better Understanding of Obstructive Sleep Apnea Treatment Effects on Cardiovascular Disease Outcomes through Machine Learning Approaches: A Narrative Review

Obstructive sleep apnea (OSA) affects almost a billion people worldwide and is associated with a myriad of adverse health outcomes. Among the most prevalent and morbid are cardiovascular diseases (CVDs). Nonetheless, randomized controlled trials (RCTs) of OSA treatment have failed to show improvements in CVD outcomes. A major limitation in our field is the lack of precision in defining OSA and specifically subgroups with the potential to benefit from therapy. Further, this has called into question the validity of using the time-honored apnea-hypopnea index as the ultimate defining criteria for OSA. Recent applications of advanced statistical methods and machine learning have brought to light a variety of OSA endotypes and phenotypes. These methods also provide an opportunity to understand the interaction between OSA and comorbid diseases for better CVD risk stratification. Lastly, machine learning and specifically heterogeneous treatment effects modeling can help uncover subgroups with differential outcomes after treatment initiation. In an era of data sharing and big data, these techniques will be at the forefront of OSA research. Advanced data science methods, such as machine-learning analyses and artificial intelligence, will improve our ability to determine the unique influence of OSA on CVD outcomes and ultimately allow us to better determine precision medicine approaches in OSA patients for CVD risk reduction. In this narrative review, we will highlight how team science via machine learning and artificial intelligence applied to existing clinical data, polysomnography, proteomics, and imaging can do just that.

Predicting obstructive sleep apnea hypopnea syndrome using three-dimensional optical devices: A systematic review

Purpose

As a global health concern, the diagnosis of obstructive sleep apnea hypopnea syndrome (OSAHS), characterized by partial reductions and complete pauses in ventilation, has garnered significant scientific and public attention. With the advancement of digital technology, the utilization of three-dimensional (3D) optical devices demonstrates unparalleled potential in diagnosing OSAHS. This study aimed to review the current literature to assess the accuracy of 3D optical devices in identifying the prevalence and severity of OSAHS.

Methods

A systematic literature search was conducted in the Web of Science, Scopus, PubMed/MEDLINE, and Cochrane Library databases for English studies published up to April 2024. Peer-reviewed researches assessing the diagnostic utility of 3D optical devices for OSAHS were included. The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) guideline was employed to appraise the risk of bias.

Results

The search yielded 3216 results, with 10 articles meeting the inclusion criteria for this study. Selected studies utilized structured light scanners, stereophotogrammetry, and red, green, blue-depth (RGB-D) cameras. Stereophotogrammetry-based 3D optical devices exhibited promising potential in OSAHS prediction.

Conclusions

The utilization of 3D optical devices holds considerable promise for OSAHS diagnosis, offering potential improvements in accuracy, cost reduction, and time efficiency. However, further clinical data are essential to assist clinicians in the early detection of OSAHS using 3D optical devices.

Deep learning technique to detect craniofacial anatomical abnormalities concentrated on middle and anterior of face in patients with sleep apnea

OBJECTIVES

The aim of this study is to propose a deep learning-based model using craniofacial photographs for automatic obstructive sleep apnea (OSA) detection and to perform design explainability tests to investigate important craniofacial regions as well as the reliability of the method.

METHODS

Five hundred and thirty participants with suspected OSA are subjected to polysomnography. Front and profile craniofacial photographs are captured and randomly segregated into training, validation, and test sets for model development and evaluation. Photographic occlusion tests and visual observations are performed to determine regions at risk of OSA. The number of positive regions in each participant is identified and their associations with OSA is assessed.

RESULTS

The model using craniofacial photographs alone yields an accuracy of 0.884 and an area under the receiver operating characteristic curve of 0.881 (95% confidence interval, 0.839-0.922). Using the cutoff point with the maximum sum of sensitivity and specificity, the model exhibits a sensitivity of 0.905 and a specificity of 0.941. The bilateral eyes, nose, mouth and chin, pre-auricular area, and ears contribute the most to disease detection. When photographs that increase the weights of these regions are used, the performance of the model improved. Additionally, different severities of OSA become more prevalent as the number of positive craniofacial regions increases.

CONCLUSIONS

The results suggest that the deep learning-based model can extract meaningful features that are primarily concentrated in the middle and anterior regions of the face.

Self-helped detection of obstructive sleep apnea based on automated facial recognition and machine learning

PURPOSE

The diagnosis of obstructive sleep apnea (OSA) relies on time-consuming and complicated procedures which are not always readily available and may delay diagnosis. With the widespread use of artificial intelligence, we presumed that the combination of simple clinical information and imaging recognition based on facial photos may be a useful tool to screen for OSA.

METHODS

We recruited consecutive subjects suspected of OSA who had received sleep examination and photographing. Sixty-eight points from 2-dimensional facial photos were labelled by automated identification. An optimized model with facial features and basic clinical information was established and tenfold cross-validation was performed. Area under the receiver operating characteristic curve (AUC) indicated the model's performance using sleep monitoring as the reference standard.

RESULTS

A total of 653 subjects (77.2% males, 55.3% OSA) were analyzed. CATBOOST was the most suitable algorithm for OSA classification with a sensitivity, specificity, accuracy, and AUC of 0.75, 0.66, 0.71, and 0.76 respectively (P < 0.05), which was better than STOP-Bang questionnaire, NoSAS scores, and Epworth scale. Witnessed apnea by sleep partner was the most powerful variable, followed by body mass index, neck circumference, facial parameters, and hypertension. The model's performance became more robust with a sensitivity of 0.94, for patients with frequent supine sleep apnea.

CONCLUSION

The findings suggest that craniofacial features extracted from 2-dimensional frontal photos, especially in the mandibular segment, have the potential to become predictors of OSA in the Chinese population. Machine learning-derived automatic recognition may facilitate the self-help screening for OSA in a quick, radiation-free, and repeatable manner.

Significance of Plasma Irisin, Adiponectin, and Retinol Binding Protein-4 Levels as Biomarkers for Obstructive Sleep Apnea Syndrome Severity

OBJECTIVE

Obstructive sleep apnea syndrome (OSAS) is a common sleep disorder that is caused by the reduction or cessation of airflow in the upper airway. Irisin, retinol-binding protein-4 (RBP-4), and adiponectin are the three significant factors in the metabolic process of the human body. The objective of this study was to investigate whether plasma irisin, RBP-4, and adiponectin levels are associated with the severity of OSAS.

METHODS

According to inclusion and exclusion criteria, 125 patients with OSAS and 46 healthy, gender-matched controls were included in this study. The patients were classified according to the apnea hypopnea index (AHI) as 14 mild cases (5 < AHI < 15), 23 moderate OSAS cases (15 < AHI < 30), and 88 severe OSAS cases (AHI > 30). The plasma irisin, RBP-4, and adiponectin levels were measured and compared between groups.

RESULTS

RBP-4 levels were higher in severe OSAS compared to other groups, and irisin levels were significantly lower in severe OSAS compared to other groups. There was a negative correlation between irisin and RBP-4 (r = -0.421; p < 0.001), and irisin and AHI (r = -0.834; p < 0.001), and a positive correlation between irisin and adiponectin (r = 0.240; p = 0.002). There was a negative correlation between RBP-4 and adiponectin (r = -0.507; p < 0.001) and a positive correlation between RBP-4 and AHI (r = 0.473; p < 0.001). As a predictor of OSAS, adiponectin showed the highest specificity (84.8%) and RBP-4 the highest sensitivity (92.0%).

CONCLUSION

Circulating adiponectin, irisin, and RBP-4 may be new biomarkers in OSAS patients in addition to risk factors such as diabetes, obesity, and hypertension. When polysomnography is not available, these parameters and clinical data can be used to diagnose the disease. As a result, patients with an AHI score greater than thirty should be closely monitored for metabolic abnormalities.

Geometric morphometrics and machine learning from three-dimensional facial scans for difficult mask ventilation prediction

Background

Unanticipated difficult mask ventilation (DMV) is a potentially life-threatening event in anesthesia. Nevertheless, predicting DMV currently remains a challenge. This study aimed to verify whether three dimensional (3D) facial scans could predict DMV in patients scheduled for general anesthesia.

Methods

The 3D facial scans were taken on 669 adult patients scheduled for elective surgery under general anesthesia. Clinical variables currently used as predictors of DMV were also collected. The DMV was defined as the inability to provide adequate and stable ventilation. Spatially dense landmarks were digitized on 3D scans to describe sufficient details for facial features and then processed by 3D geometric morphometrics. Ten different machine learning (ML) algorithms, varying from simple to more advanced, were introduced. The performance of ML models for DMV prediction was compared with that of the DIFFMASK score. The area under the receiver operating characteristic curves (AUC) with its 95% confidence interval (95% CI) as well as the specificity and sensitivity were used to evaluate the predictive value of the model.

Results

The incidence of DMV was 35/669 (5.23%). The logistic regression (LR) model performed best among the 10 ML models. The AUC of the LR model was 0.825 (95% CI, 0.765-0.885). The sensitivity and specificity of the model were 0.829 (95% CI, 0.629-0.914) and 0.733 (95% CI, 0.532-0.819), respectively. The LR model demonstrated better predictive performance than the DIFFMASK score, which obtained an AUC of 0.785 (95% CI, 0.710-0.860) and a sensitivity of 0.686 (95% CI, 0.578-0.847). Notably, we identified a significant morphological difference in the mandibular region between the DMV group and the easy mask ventilation group.

Conclusion

Our study indicated a distinct morphological difference in the mandibular region between the DMV group and the easy mask ventilation group. 3D geometric morphometrics with ML could be a rapid, efficient, and non-invasive tool for DMV prediction to improve anesthesia safety.

Do brachycephaly and nose size predict the severity of obstructive sleep apnea (OSA)? A sample-based geometric morphometric analysis of craniofacial variation in relation to OSA syndrome and the role of confounding factors

Obstructive sleep apnea is a common disorder that leads to sleep fragmentation and is potentially bidirectionally related to a variety of comorbidities, including an increased risk of heart failure and stroke. It is often considered a consequence of anatomical abnormalities, especially in the head and neck, but its pathophysiology is likely to be multifactorial in origin. With geometric morphometrics, and a large sample of adults from the Study for Health in Pomerania, we explore the association of craniofacial morphology to the apnea-hypopnea index used as an estimate of obstructive sleep apnea severity. We show that craniofacial size and asymmetry, an aspect of morphological variation seldom analysed in obstructive sleep apnea research, are both uncorrelated to apnea-hypopnea index. In contrast, as in previous analyses, we find evidence that brachycephaly and larger nasal proportions might be associated to obstructive sleep apnea severity. However, this correlational signal is weak and completely disappears when age-related shape variation is statistically controlled for. Our findings suggest that previous work might need to be re-evaluated, and urge researchers to take into account the role of confounders to avoid potentially spurious findings in association studies.