Deep-learning for seizure forecasting in canines with epilepsy

Abstract

Objective. This paper introduces a fully automated, subject-specific deep-learning convolutional neural network (CNN) system for forecasting seizures using ambulatory intracranial EEG (iEEG). The system was tested on a hand-held device (Mayo Epilepsy Assist Device) in a pseudo-prospective mode using iEEG from four canines with naturally occurring epilepsy. Approach. The system was trained and tested on 75 seizures collected over 1608 d utilizing a genetic algorithm to optimize forecasting hyper-parameters (prediction horizon (PH), median filter window length, and probability threshold) for each subject-specific seizure forecasting model. The trained CNN models were deployed on a hand-held tablet computer and tested on testing iEEG datasets from four canines. The results from the iEEG testing datasets were compared with Monte Carlo simulations using a Poisson random predictor with equal time in warning to evaluate seizure forecasting performance. Main results. The results show the CNN models forecasted seizures at rates significantly above chance in all four dogs (p < 0.01, with mean 0.79 sensitivity and 18% time in warning). The deep learning method presented here surpassed the performance of previously reported methods using computationally expensive features with standard machine learning methods like logistic regression and support vector machine classifiers. Significance. Our findings principally support the feasibility of deploying trained CNN models on a hand-held computational device (Mayo Epilepsy Assist Device) that analyzes streaming iEEG data for real-time seizure forecasting.

Seizure forecasting and cyclic control of seizures

“…The NeuroVista device predicted patients’ seizures with 65%‐100% sensitivity. Similar advancements using implantable seizure forecasting devices were also made in naturally occurring canine epilepsy 32,33 . In 2014, two large‐scale crowdsourcing competitions were launched to allow the machine learning community to tackle seizure‐detection and seizure‐forecasting algorithm development, 17,34 which paved the way for a subsequent competition 35 .…”

Deep Convolutional Neural Network Based Interictal-Preictal Electroencephalography Prediction: Application to Focal Cortical Dysplasia Type-II

“…However, they reported highly varied distributions of the HFOs among patients and poor performance outcomes in their seizure predictors. Recent studies using implantable devices for seizure detection and prediction acquired intracranial EEG recordings at sampling frequencies lower than 400 Hz in human ( 29 – 32 , 35 ) and canine ( 37 ) subjects, likely to lower power consumption and increase data processing efficiency. Our results suggest that sampling frequencies from 128 to 512 Hz have no significant impact on interictal-preictal discriminability by CNN-based classification models.…”

“…The band-pass filtered signals were segmented into 30 s epochs varying the amount of overlap based on the preictal lengths to equalize the number of preictal and interictal epochs as close as possible. Each epoch was transformed to a time-frequency two-dimensional data matrix by the short-time Fourier transform using a 1 s Hamming window with 50% overlap to capture non-stationary EEG characteristics both in time and frequency domains as seen in recent human ( 6 , 35 ) and canine ( 37 ) studies on the deep learning-based seizure prediction. Coefficients between 55 and 64 Hz were excluded from the time-frequency data matrix down-sampled at 256–512 Hz to remove power line noise effects.…”

Detection of Epileptic Seizure Using Pretrained Deep Convolutional Neural Network and Transfer Learning

“…It is more preferable to do feature extraction before classification than to train the classifier model directly with raw EEG samples. However, in recent studies for the detection of epilepsy, raw EEG data were used in deep machine learning models without feature extraction [68-70].…”