Detecting events in physiological time series

Detecting events in physiological time series

Problem being addressed

Semi-automatic detection of inter-ictal events in EEG recorded inside MRI using deep learning

Solution

The authors propose a deep learning based detection method to tackle this problem. First, they train a ResNet neural network model by mapping EEG data into a space in which the same types of events lie close to each other but far from other event types or from baseline. A multi-task learning strategy is used to simultaneously classify different types of events and learn a Euclidean distance embedding space. Then, they get candidate events by applying the model to EEG of new studies, let the expert edit them, and apply event detection algorithm using similarities in the output layer space

Advantages of this solution

Provides recommended events and reduces costly manual marking Shows improvement in detectability and ROC curves compared to standard template matching techniques

Solution originally applied in these industries

healthcare

Healthcare Sector

Possible New Application of the Work

/static/media/aerospaceAndDefence-1.faaa8901.png

Aerospace & Defence Sector

Detecting unique radar events.

/static/media/agriculture-1.a14ab412.png

Agriculture Industry

Detecting abnormalities measured by environmental sensors

/static/media/healthcare.4e8e39b0.png

Healthcare Sector

detecting physiological time series abnormalities in ECG, SPO2, HRV and other measures

Author of original research described in this blitzcard:

question-mark-icon

Name of the author who conducted the original research that this blitzcard is based on.

Source DOI: #############check-icon

Source URL: #############check-icon


search-iconBrowse all blitzcards