As a result, for example, a variable number of OSA patient clusters (3 to 7) 15, 16, 17, 18 and asthma patient clusters (3 to 6) were chosen in previous studies 19, 20, 21, 22. However, the clustering algorithms used in these studies require the number of clusters to be manually determined either before (e.g. This has uncovered new phenotypes of complex and heterogeneous diseases including not only OSA 15, 16, 17, 18 but also other diseases such as asthma 19, 20, 21, 22, chronic obstructive pulmonary disease 23, 24, chronic heart failure 25, sepsis 26, Parkinson’s disease 27, and diabetes 28, 29. To analyze multidimensional clinical data such as PSG, cluster analysis has been widely used. This indicates that the AHI alone is not enough to explain the comorbidity developments in OSA patients, and highlights the need for OSA phenotyping based on all PSG data. Interestingly, this cluster had significantly higher risks of cardiovascular diseases among clusters with a mild degree of AHI. This identified a cluster of patients with high periodic limb movements (PLM), which is an important PSG feature. For example, the K-means algorithm was used to cluster OSA patients into seven phenotypes based on their PSG 17. graph-based clustering 15 and K-means clustering 16, 17), which is well-suited for OSA phenotyping because PSG data is multidimensional data containing various information regarding the patient’s sleep generated during their initial visit for diagnosis. Recently, more comprehensive OSA phenotypes have been identified through clustering PSG data (e.g. This indicates the need for more comprehensive phenotyping of OSA beyond just the AHI (Supplementary Table S1) 14. For instance, sole AHI fails to discriminate a patient’s comorbidity outcomes within the same AHI severity 10, 11, 12, 13. The AHI alone is likely an over-simplistic index to explain the heterogeneity and complexity of the disease. The prognosis of associated diseases is also made primarily based on the AHI 10. This is also used to classify patients into phenotypes such as mild, moderate, and severe OSA (Supplementary Table S1). In particular, only the apnea–hypopnea index (AHI), which is the number of apneas (temporary cessation of breathing) and hypopneas (partial blockage of the airway) per hour of sleep, is used. However, only the respiratory events are used for the conventional diagnosis of OSA. The standard test for diagnosing OSA 8 is based on the polysomnography (PSG) 9, which records various parameters such as sleep architecture, respiratory events, oxygen desaturation, and limb movements during sleep. Obstructive sleep apnea (OSA) is one of the most common sleep disorders 1 and is a risk factor of various diseases, including cardiovascular, neurovascular, and metabolic diseases 2, 3, 4, 5, 6, 7. The phenotyping framework based on the combination of unsupervised and supervised machine learning methods can also be applied to other complex, heterogeneous diseases for phenotyping patients and identifying important features for high-risk phenotypes. These results can also be used to automatically phenotype new patients and predict their comorbidity risks solely based on their PSG data. Furthermore, the key features of highly comorbid phenotypes were identified through supervised learning rather than subjective choice. Importantly, these new phenotypes show statistically different comorbidity development for OSA-related cardio-neuro-metabolic diseases, unlike the conventional single-metric apnea–hypopnea index-based phenotypes. This clusters 2277 OSA patients to six phenotypes based on their multidimensional polysomnography (PSG) data. Here, to minimize such subjective decisions for highly confident phenotyping, we develop a multimetric phenotyping framework by combining supervised and unsupervised machine learning. However, the number of clusters and key phenotypic features have been subjectively selected, reducing the reliability of the phenotyping results. Unsupervised clustering models have been widely used for multimetric phenotyping of complex and heterogeneous diseases such as diabetes and obstructive sleep apnea (OSA) to more precisely characterize the disease beyond simplistic conventional diagnosis standards.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |