|
|
 |
| |
|
|
|
Diagnosis of Severe Obstructive Sleep Apnea with Model Designed using Genetic Algorithm and Ensemble Support Vector Machine |
|
|
|
PP: 149-157 |
|
|
|
Author(s) |
|
|
|
Liang-Wen Hang,
Hsuan-Hung Lin,
Chung-Chuan Cheng,
John Y. Chiang,
Hsiang-Ling Wang,
Fu Chen,
|
|
|
|
Abstract |
|
|
| Obstructive sleep apnea (OSA) is a general sleep disorder and is a significant cause of motor vehicle crashes and chronic
diseases. The severity of the respiratory events is measured by the frequency and duration of apneas and hypopneas per hour of sleep,
namely apnea-hypopnea index (AHI), using polysomnography (PSG). Suspected patients can be classified as normal (AHI<5), mild
(5≤AHI<15), moderate (15≤AHI<30), and severe (AHI≥30). Although PSG is treated as the gold standard for the diagnosis of OSA,
its shortcoming includes technical expertise is required and timely access is restricted. Thus, home pulse oximetry has been proposed
as a valuable and effective tool for screening patients with OSA. Support vector machine (SVM) is believed to be more efficient
than neural network and traditional statistical-based classifiers. Nonetheless, it is critical to determine suitable parameters to increase
classification performance. Furthermore, an ensemble of SVM classifiers use multiple models to obtain better predictive accuracy and
are more stable than models consist of a single model. Genetic algorithm (GA), on the other hand, is able to find optimal solution
within an acceptable time, and is faster than dynamic programming with exhaustive searching strategy. By taking the advantage of
GA in quickly selecting the salient features and adjusting SVM parameters, it was combined with ensemble SVM to design a clinical
decision support system (CDSS) for the diagnosis of patients with severe OSA, and then followed by PSG to further discriminate
normal mild and moderate patients. The results show that ensemble SVM classifiers demonstrate better diagnosing performance than
models consisting of a single SVM model and logistic regression analysis. Additionally, the oximetry/PSG diagnostic scheme was
shown to have higher cost-effectiveness in the diagnosis of OSA patients with an average cost ratio of 0.66 and an average waiting time
ratio of 0.40 compared to the traditional scheme with PSG examination only. |
|
|
|
|
|
|
|
