Classification of Epileptic EEG Signals with Stacked Sparse Autoencoder Based on Deep Learning

Qin, Lin; Ye, Shu-qun; Huang, Xiu-mei; Li, Si-you; Zhang, Meizhen; Xue, Yun; Chen, Wensheng

doi:10.1007/978-3-319-42297-8_74

Cited by 48 publications

(20 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SAE has often been used for unsupervised EHR phenotyping 83 or sparse EEG feature representation. 79 , 101 , 102 For DAE, the reconstruction is based on randomly corrupted inputs, through which the model gains robustness against missing data or noise. DAE has been used for learning robust representations of human physiology, 10 , 30 , 82 deriving robust patient representation from EHRs, 30 or extracting EHR phenotypes that can be paired with genetic data to identify disease-gene associations.…”

Section: Resultsmentioning

confidence: 99%

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Xiao¹,

Choi

Sun

2018

Journal of the American Medical Informatics Association

610

346

View full text Add to dashboard Cite

ObjectiveTo conduct a systematic review of deep learning models for electronic health record (EHR) data, and illustrate various deep learning architectures for analyzing different data sources and their target applications. We also highlight ongoing research and identify open challenges in building deep learning models of EHRs.Design/methodWe searched PubMed and Google Scholar for papers on deep learning studies using EHR data published between January 1, 2010, and January 31, 2018. We summarize them according to these axes: types of analytics tasks, types of deep learning model architectures, special challenges arising from health data and tasks and their potential solutions, as well as evaluation strategies.ResultsWe surveyed and analyzed multiple aspects of the 98 articles we found and identified the following analytics tasks: disease detection/classification, sequential prediction of clinical events, concept embedding, data augmentation, and EHR data privacy. We then studied how deep architectures were applied to these tasks. We also discussed some special challenges arising from modeling EHR data and reviewed a few popular approaches. Finally, we summarized how performance evaluations were conducted for each task.DiscussionDespite the early success in using deep learning for health analytics applications, there still exist a number of issues to be addressed. We discuss them in detail including data and label availability, the interpretability and transparency of the model, and ease of deployment.

show abstract

Section: Resultsmentioning

confidence: 99%

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Xiao¹,

Choi

Sun

2018

Journal of the American Medical Informatics Association

610

346

View full text Add to dashboard Cite

show abstract

“…Lin et al 126 and Viyaratne et al, 127 using deep learning, implemented automatic detection of epileptic seizures by learning the complex and nonstationary epileptic EEG signals and performed feature extraction without relying on methods that are supervised and require domain-specific expertise. The proposal is capable of learning more abstract and high-level representations, which allows discovering significant differences between seizure and normal EEG signals.…”

Section: Epilepsy Applications Using Brain Electrical Activity and Dementioning

confidence: 99%

Automated analysis of seizure semiology and brain electrical activity in presurgery evaluation of epilepsy: A focused survey

et al. 2017

View full text Add to dashboard Cite

Epilepsy being one of the most prevalent neurological disorders, affecting approximately 50 million people worldwide, and with almost 30-40% of patients experiencing partial epilepsy being nonresponsive to medication, epilepsy surgery is widely accepted as an effective therapeutic option. Presurgical evaluation has advanced significantly using noninvasive techniques based on video monitoring, neuroimaging, and electrophysiological and neuropsychological tests; however, certain clinical settings call for invasive intracranial recordings such as stereoelectroencephalography (SEEG), aiming to accurately map the eloquent brain networks involved during a seizure. Most of the current presurgical evaluation procedures focus on semiautomatic techniques, where surgery diagnosis relies immensely on neurologists' experience and their time-consuming subjective interpretation of semiology or the manifestations of epilepsy and their correlation with the brain's electrical activity. Because surgery misdiagnosis reaches a rate of 30%, and more than one-third of all epilepsies are poorly understood, there is an evident keen interest in improving diagnostic precision using computer-based methodologies that in the past few years have shown near-human performance. Among them, deep learning has excelled in many biological and medical applications, but has advanced insufficiently in epilepsy evaluation and automated understanding of neural bases of semiology. In this paper, we systematically review the automatic applications in epilepsy for human motion analysis, brain electrical activity, and the anatomoelectroclinical correlation to attribute anatomical localization of the epileptogenic network to distinctive epilepsy patterns. Notably, recent advances in deep learning techniques will be investigated in the contexts of epilepsy to address the challenges exhibited by traditional machine learning techniques. Finally, we discuss and propose future research on epilepsy surgery assessment that can jointly learn across visually observed semiologic patterns and recorded brain electrical activity.

show abstract

“…Third, supervised fine-tuning is used to improve the deep neural network performance using backpropagation on the whole multilayer network. More details of the SSAE can be found elsewhere [27]- [29]. …”

Section: Supervised Classification-stacked Sparse Auto-encodersmentioning

confidence: 99%

A fully automatic deep learning method for atrial scarring segmentation from late gadolinium-enhanced MRI images

Yang

Zhuang

Khan

et al. 2017

2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017)

View full text Add to dashboard Cite

Precise and objective segmentation of atrial scarring (SAS) is a prerequisite for quantitative assessment of atrial fibrillation using non-invasive late gadolinium-enhanced (LGE) MRI. This also requires accurate delineation of the left atrium (LA) and pulmonary veins (PVs) geometry. Most previous studies have relied on manual segmentation of LA wall and PVs, which is a tedious and error-prone procedure with limited reproducibility. There are many attempts on automatic SAS using simple thresholding, histogram analysis, clustering and graph-cut based approaches; however, in general, these methods are considered as unsupervised learning thus subject to limited segmentation accuracy. In this study, we present a fully-automated multi-atlas based whole heart segmentation method to derive the LA and PVs geometry objectively that is followed by a fully automatic deep learning method for SAS. Our deep learning method consists of a feature extraction step via super-pixel over-segmentation and a supervised classification step via stacked sparse auto-encoders. We demonstrate the efficacy of our method on 20 clinical LGE MRI scans acquired from a longstanding persistent atrial fibrillation cohort. Both quantitative and qualitative results show that our fully automatic method obtained accurate segmentation results compared to the manual segmentation based ground truths. INTRODUCTIONAtrial fibrillation (AF) is the most common sustained heart rhythm disturbance encountered in adult cardiology. Several studies have shown that AF is correlated with electrical, contractile, and structural remodeling in the left atrium (LA) [1]. Moreover, LA fibrosis may be arrhythmogenic that causes more aggressive symptoms and makes difficulties in the management of AF [1]. Minimally invasive catheter ablation (CA) using radio-frequency energy has become one of the most common treatments for AF patients refractory to drug treatment [2]. CA aims to electrically isolate the pulmonary veins (PVs) from the left atrial (LA) body because previous studies show that ectopic beats from the PVs can frequently trigger the AF [3]. In this context, techniques have been developed to evaluate the LA wall composition and assess the circumferential PVs scarring that results from CA in order to understand the AF with proper management and prognosis. At present, electro-anatomical mapping (EAM) system, which is performed during the electrophysiological study, is considered to be a clinical reference standard technique for the assessment of the LA substrate and ablation-induced scarring. However, EAM is invasive and suffers from ionizing radiation and its suboptimal accuracy, which has reported errors of up to 10 mm in the localization of scar tissue [4]. Noninvasive late gadolinium-enhanced (LGE) MRI is an established method for visualizing and assessing myocardial infarction or fibrosis [5]. This is ascribed to the altered wash-in and wash-out contrast agent kinetics, and the hyper-enhancement reflects the increased interstitial space of the myocardium wit...

show abstract

Classification of Epileptic EEG Signals with Stacked Sparse Autoencoder Based on Deep Learning

Cited by 48 publications

References 10 publications

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Automated analysis of seizure semiology and brain electrical activity in presurgery evaluation of epilepsy: A focused survey

A fully automatic deep learning method for atrial scarring segmentation from late gadolinium-enhanced MRI images

Contact Info

Product

Resources

About