Department of Neurology, Weill Cornell Medicine, 525 E. 68(th) St F-610, New York, NY 10065, United States; Department of Health Policy & Research, Weill Cornell Medicine, 402 E. 67(th) St, New York, NY 10065, United States.
Department of Health Policy & Research, Weill Cornell Medicine, 402 E. 67(th) St, New York, NY 10065, United States.
Seizure. 2019 Oct;71:124-131. doi: 10.1016/j.seizure.2019.07.009. Epub 2019 Jul 8.
Non-convulsive seizures are common in critically ill patients, and delays in diagnosis contribute to increased morbidity and mortality. Many intensive care units employ continuous EEG (cEEG) for seizure monitoring. Although cEEG is continuously recorded, it is often reviewed intermittently, which may delay seizure diagnosis and treatment. This may be mitigated with automated seizure detection. In this study, we develop and evaluate convolutional neural networks (CNN) to automate seizure detection on EEG spectrograms.
Adult EEGs (12 patients, 12 EEGs, 33 seizures) from New-York Presbyterian Hospital (NYP) and pediatric EEGs (22 patients, 130 EEGs, 177 seizures) from Children's Hospital Boston (CHB) were converted into spectrograms. To simulate a telemetry display, seizure and non-seizure events on spectrograms were sequentially sampled as images across a detection window (26,380 total images). Four CNN models of increasing complexity (number of layers) were trained, cross-validated, and tested on CHB and NYP spectrographic images. All CNNs were based on the VGG-net architecture, with adjustments to alleviate overfitting.
For spectrographically visible seizures, two CNN models (containing 4 and 7 convolution layers) achieved >90% seizure detection sensitivity and specificity on the CHB test set and >90% sensitivity and 75-80% specificity on the NYP test set. The one CNN model (10 convolution layers) did not converge during training; while another CNN (2 convolution layers) performed poorly (60% sensitivity and 32% specificity) on the NYP test set.
Seizure detection on EEG spectrograms with CNN models is feasible with sensitivity and specificity potentially suitable for clinical use.
非惊厥性发作在危重症患者中很常见，诊断延迟会导致发病率和死亡率增加。许多重症监护病房采用连续脑电图（cEEG）进行癫痫监测。尽管 cEEG 是连续记录的，但通常是间歇性地进行复查，这可能会延迟癫痫的诊断和治疗。这可以通过自动化癫痫检测来缓解。在这项研究中，我们开发并评估了卷积神经网络（CNN），以实现脑电图频谱图上的癫痫自动检测。
将来自纽约长老会医院（NYP）的成人脑电图（12 例患者，12 个脑电图，33 个癫痫发作）和波士顿儿童医院（CHB）的儿科脑电图（22 例患者，130 个脑电图，177 个癫痫发作）转换为频谱图。为了模拟遥测显示，在检测窗口（共 26,380 个总图像）中，按顺序对频谱图上的癫痫发作和非癫痫发作事件进行采样作为图像。对四个复杂度（层数）递增的 CNN 模型进行了训练、交叉验证和测试，测试集来自 CHB 和 NYP 光谱图像。所有的 CNN 都基于 VGG-net 架构，并进行了调整以减轻过拟合。
对于在频谱图上可见的癫痫发作，两个 CNN 模型（包含 4 层和 7 层卷积）在 CHB 测试集上达到了>90%的癫痫检测灵敏度和特异性，在 NYP 测试集上达到了>90%的灵敏度和 75-80%的特异性。一个 CNN 模型（10 层卷积）在训练过程中没有收敛；而另一个 CNN（2 层卷积）在 NYP 测试集上表现不佳（灵敏度 60%，特异性 32%）。
使用 CNN 模型对脑电图频谱图进行癫痫检测是可行的，其灵敏度和特异性可能适合临床应用。
