Improving Brain E-Health Services via High-Performance EEG Classification With Grouping Bayesian Optimization

Ke, Hengjin; Chen, Dan; Shi, Benyun; Zhang, Jindong; Liu, Xianzeng; Zhang, Xinhua; Li, Xiaoli

doi:10.1109/tsc.2019.2962673

Cited by 54 publications

(31 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with the relatively large datasets used in previous studies, our dataset is a multi-labeling corpus and is small and highly imbalanced with respect to four out of the five disease labels. Second, we employed a cost-sensitive learning approach with the same weights across all models to address the imbalance issue because this approach has been demonstrated to be effective for addressing the imbalance problem in the medical domain ( 32 ) and several multi-class datasets with varying levels of imbalance ( 33 ). However, the best weights may vary depending on the network architectures and the employed pre-trained techniques.…”

Section: Discussionmentioning

confidence: 99%

“…However, the best weights may vary depending on the network architectures and the employed pre-trained techniques. Hence, improved results may be obtained experimentally by using a hyper-parameter search ( 34 ). Furthermore, in addition to cost-sensitive learning, a variety of methods, such as oversampling and data augmentation, are available to address imbalance problems ( 23 , 35 ).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based Natural Language Processing for Screening Psychiatric Patients

Dai

Lee

et al. 2021

Front. Psychiatry

View full text Add to dashboard Cite

The introduction of pre-trained language models in natural language processing (NLP) based on deep learning and the availability of electronic health records (EHRs) presents a great opportunity to transfer the “knowledge” learned from data in the general domain to enable the analysis of unstructured textual data in clinical domains. This study explored the feasibility of applying NLP to a small EHR dataset to investigate the power of transfer learning to facilitate the process of patient screening in psychiatry. A total of 500 patients were randomly selected from a medical center database. Three annotators with clinical experience reviewed the notes to make diagnoses for major/minor depression, bipolar disorder, schizophrenia, and dementia to form a small and highly imbalanced corpus. Several state-of-the-art NLP methods based on deep learning along with pre-trained models based on shallow or deep transfer learning were adapted to develop models to classify the aforementioned diseases. We hypothesized that the models that rely on transferred knowledge would be expected to outperform the models learned from scratch. The experimental results demonstrated that the models with the pre-trained techniques outperformed the models without transferred knowledge by micro-avg. and macro-avg. F-scores of 0.11 and 0.28, respectively. Our results also suggested that the use of the feature dependency strategy to build multi-labeling models instead of problem transformation is superior considering its higher performance and simplicity in the training process.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Deep Learning-Based Natural Language Processing for Screening Psychiatric Patients

Dai

Lee

et al. 2021

Front. Psychiatry

View full text Add to dashboard Cite

show abstract

“…To achieve the performance improvement as discussed above, a many-task computing approach with a GeoSOT-encoded quad task-tree is proposed to tackle the incredible large-scale data clustering challenge of a million-scale of high-dimensional fire points. As a matter of fact, this study essentially extended the idea of big data factorization proposed by Chen et al [44][45][46] to employ a GeoSOT global division method so as to gradually break down the small grid tasks. Eventually, a GeoSOT-encoded quad task tree could be formed from these small bags of grid tasks for a fully exploiting of the parallelism.…”

Section: Discussionmentioning

confidence: 99%

Spatial-Temporal Distribution Analysis of Industrial Heat Sources in the US with Geocoded, Tree-Based, Large-Scale Clustering

Liu

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

Heavy industrial burning contributes significantly to the greenhouse gas (GHG) emissions. It is responsible for almost one-quarter of the global energy-related CO2 emissions and its share continues to grow. Mostly, those industrial emissions are accompanied by a great deal of high-temperature heat emissions from the combustion of carbon-based fuels by steel, petrochemical, or cement plants. Fortunately, these industrial heat emission sources treated as thermal anomalies can be detected by satellite-borne sensors in a quantitive way. However, most of the dominant remote sensing-based fire detection methods barely work well for heavy industrial heat source discernment. Although the object-oriented approach, especially the data clustering-based approach, has guided a novel method of detection, it is still limited by the costly computation and storage resources. Furthermore, when scaling to a national, or even global, long time-series detection, it is greatly challenged by the tremendous computation introduced by the incredible large-scale data clustering of tens of millions of high-dimensional fire data points. Therefore, we proposed an improved parallel identification method with geocoded, task-tree-based, large-scale clustering for the spatial-temporal distribution analysis of industrial heat emitters across the United States from long time-series active Visible Infrared Imaging Radiometer Suite (VIIRS) data. A recursive k-means clustering method is introduced to gradually segment and cluster industrial heat objects. Furthermore, in order to avoid the blindness caused by random cluster center initialization, the time series VIIRS hotspots data are spatially pre-grouped into GeoSOT-encoded grid tasks which are also treated as initial clustering objects. In addition, some grouped parallel clustering strategy together with geocoding-aware task tree scheduling is adopted to sufficiently exploit parallelism and performance optimization. Then, the spatial-temporal distribution pattern and its changing trend of industrial heat emitters across the United States are analyzed with the identified industrial heat sources. Eventually, the performance experiment also demonstrated the efficiency and encouraging scalability of this approach.

show abstract

“…Electroencephalogram (EEG) (31,32), event-related potential (ERP) (33), magnetic resonance imaging (MRI) (34), functional magnetic resonance imaging (fMRI) (35), and positron emission tomography (PET) (36) are imaging tools that are commonly used in psychiatric research. These techniques generate images that are either static (MRI) or time-varying (EEG, ERP, fMRI, PET).…”

Section: Neuroimaging Datamentioning

confidence: 99%

Illuminating the Black Box: Interpreting Deep Neural Network Models for Psychiatric Research

Sheu

2020

Front. Psychiatry

View full text Add to dashboard Cite

Psychiatric research is often confronted with complex abstractions and dynamics that are not readily accessible or well-defined to our perception and measurements, making data-driven methods an appealing approach. Deep neural networks (DNNs) are capable of automatically learning abstractions in the data that can be entirely novel and have demonstrated superior performance over classical machine learning models across a range of tasks and, therefore, serve as a promising tool for making new discoveries in psychiatry. A key concern for the wider application of DNNs is their reputation as a "black box" approach-i.e., they are said to lack transparency or interpretability of how input data are transformed to model outputs. In fact, several existing and emerging tools are providing improvements in interpretability. However, most reviews of interpretability for DNNs focus on theoretical and/or engineering perspectives. This article reviews approaches to DNN interpretability issues that may be relevant to their application in psychiatric research and practice. It describes a framework for understanding these methods, reviews the conceptual basis of specific methods and their potential limitations, and discusses prospects for their implementation and future directions.

show abstract

Improving Brain E-Health Services via High-Performance EEG Classification With Grouping Bayesian Optimization

Cited by 54 publications

References 31 publications

Deep Learning-Based Natural Language Processing for Screening Psychiatric Patients

Deep Learning-Based Natural Language Processing for Screening Psychiatric Patients

Spatial-Temporal Distribution Analysis of Industrial Heat Sources in the US with Geocoded, Tree-Based, Large-Scale Clustering

Illuminating the Black Box: Interpreting Deep Neural Network Models for Psychiatric Research

Contact Info

Product

Resources

About