New tool merges expert knowledge and deep learning features to detect sleep apnea


UNIVERSITY PARK, Pa .– People who suspect they have sleep apnea – and the doctors who diagnose it – may soon have a more effective way to automatically detect the disease at home, thanks to a new method developed by researchers at Penn State College of Information Science and Technology.

The new tool, which the researchers say surpasses all existing core methods, combines deep learning technology with expert knowledge. It automatically learns patterns from electrocardiograph (ECG) data collected by home devices, making it a faster and more ideal solution than other sleep apnea diagnostics.

“The current standard approach to detecting sleep apnea is for a patient to stay in the hospital overnight to record a polysomnography (sleep study) under the supervision of a clinical practitioner,” said Guanjie Huang, doctoral student in information science and technology and main author. On paper. “The process is long, tedious, intrusive and late. “

Huang explained that after collecting a patient’s data through a sleep study, which measures brain waves, blood oxygen levels, heart rate, breathing and body movements, clinicians should then devote more time and resources to analyze it.

“It is essential to design an accurate model to automatically analyze the data and help doctors quickly detect sleep apnea,” said Huang.

Other tools for automatically detecting sleep apnea via home devices exist using computer models built either through traditional machine learning methods, which draw on expert human knowledge to design created features. by hand that can identify sleep apnea conditions in a dataset, or via deep learning methods, which eliminate the need for such experts due to immense amounts of data. But, according to Huang, there are limits to these stand-alone approaches.

“The traditional machine learning method usually only needs a small amount of data to learn a robust classifier, but it requires a careful process of feature extraction and selection,” Huang explained. “The deep learning method usually results in better performance, but requires a large data set. “

Huang’s model, called ConCAD (Contrastive Learning-based Cross Attention for Sleep Apnea Detection), simultaneously leverages deep learning features and the specialist knowledge of traditional machine learning to better detect sleep apnea. The model specifically draws on expert knowledge of the RR interval and peak R envelope – existing methods to detect sleep apnea by measuring the intervals between and peak R wave, which measure the heart rate in a patient’s ventricular walls, in a standard ECG. ConCAD uses a cross-attention mechanism – a deep learning module that assigns weights to parts of each based on their importance – to merge deep learning functionality with expert knowledge functionality, putting the emphasis on ‘focusing on those that are useful and automatically ignoring those that are irrelevant.

ConCAD works by first passing the original raw ECG data through feature extractors to automatically learn patterns from expert knowledge and deep learning methods that might indicate sleep apnea. These patterns, or characteristics, are then automatically and synergistically merged and given a weight based on the important parts of each. Then, through a contrasting learning process, similar characteristics are closely associated. Finally, the data is classified according to the final characteristics of the ECG and the corresponding expert knowledge, indicating the likelihood that the patient has sleep apnea.

To test their model, the researchers used two publicly available ECG datasets containing more than 26,000 expert-annotated segments, each identifying apnea or normal sleep events. These segments included 30-second and two-and-a-half-minute entries. Compared to six existing state-of-the-art sleep apnea detection methods, ConCAD has outperformed them all. Their model accurately identified sleep apnea events 88.75% of the time in the one-minute segments and 91.22% of the time in the five-minute segments of the first data set; and 82.5% and 83.47% respectively in the second data set.

“Our results show the possibility of using ECG data for automated detection of sleep apnea, which should greatly benefit patients with sleep apnea, as they can use a personal ECG machine at home to monitor their sleep apnea conditions, “said Fenglong Ma, Assistant Professor in Information Science and Technology and Principal Investigator. In addition, the designed model can help physicians simplify the process of diagnosing sleep apnea.

Ma added, “This is a new attempt to incorporate expert knowledge into deep learning models for the detection of sleep apnea. We will continue to explore how to use expert knowledge to guide the learning of deep models. “

The researchers presented ConCAD at the European Conference on Machine Learning and the Principles and Practices of Knowledge Discovery in Databases (CELV-PKDD), held virtually from September 13 to 17.


Source link

Previous Verana Health is recruiting new talent to accelerate its growth
Next How Bosses Should Write Books

No Comment

Leave a reply

Your email address will not be published. Required fields are marked *