[论文解读] Human Intracranial EEG Quantitative Analysis and Automatic Feature Learning for Epileptic Seizure Prediction

Ramy Hussein, Mohamed Osama Ahmed+4 more|arXiv (Cornell University)|Apr 7, 2019

EEG and Brain-Computer Interfaces参考文献 28被引用 37

一句话总结

论文提出了一个CNN-based癫痫发作预测系统，使用预处理后的iEEG数据转换为图像样声谱图，并显示多尺度CNN在Kaggle iEEG数据集上优于最先进方法。

ABSTRACT

Objective: The aim of this study is to develop an efficient and reliable epileptic seizure prediction system using intracranial EEG (iEEG) data, especially for people with drug-resistant epilepsy. The prediction procedure should yield accurate results in a fast enough fashion to alert patients of impending seizures. Methods: We quantitatively analyze the human iEEG data to obtain insights into how the human brain behaves before and between epileptic seizures. We then introduce an efficient pre-processing method for reducing the data size and converting the time-series iEEG data into an image-like format that can be used as inputs to convolutional neural networks (CNNs). Further, we propose a seizure prediction algorithm that uses cooperative multi-scale CNNs for automatic feature learning of iEEG data. Results: 1) iEEG channels contain complementary information and excluding individual channels is not advisable to retain the spatial information needed for accurate prediction of epileptic seizures. 2) The traditional PCA is not a reliable method for iEEG data reduction in seizure prediction. 3) Hand-crafted iEEG features may not be suitable for reliable seizure prediction performance as the iEEG data varies between patients and over time for the same patient. 4) Seizure prediction results show that our algorithm outperforms existing methods by achieving an average sensitivity of 87.85% and AUC score of 0.84. Conclusion: Understanding how the human brain behaves before seizure attacks and far from them facilitates better designs of epileptic seizure predictors. Significance: Accurate seizure prediction algorithms can warn patients about the next seizure attack so they could avoid dangerous activities. Medications could then be administered to abort the impending seizure and minimize the risk of injury.

研究动机与目标

• Quantitatively analyze human iEEG data before and between seizures to understand brain behavior.
• Develop an efficient pre-processing pipeline to convert time-series iEEG into image-like inputs for CNNs.
• Propose a cooperative multi-scale CNN architecture to automatically learn iEEG features.
• Demonstrate seizure prediction performance across patients with drug-resistant epilepsy and compare to state-of-the-art.

提出的方法

• Downsample iEEG clips from 400 Hz to 100 Hz to reduce data size while preserving informative low-frequency content.
• Segment 10-minute clips into 1-minute segments to increase training samples.
• Apply Short-Time Fourier Transform to generate 16-channel STFT spectrogram images per segment.
• Use a multi-branch CNN with 1x1, 3x3, and 5x5 conv paths plus a parallel pooling path; concatenate outputs for FC layers.
• Train with binary cross-entropy using Adam optimizer (lr=0.001); final prediction per 10-minute clip via max over ten 1-minute predictions.
• Evaluate on sensitivity (SEN) and AUC metrics; compare against prior methods on the same Kaggle iEEG dataset.

实验结果

研究问题

• RQ1 Are interictal and preictal iEEG states statistically different enough to enable prediction across patients?
• RQ2 Can PCA reliably reduce iEEG dimensionality for seizure prediction, or is it disadvantageous?
• RQ3 Is there benefit in using all iEEG channels versus channel exclusion for seizure prediction?
• RQ4 Can a CNN learn effective local and high-level features from STFT-based iEEG spectrograms for prediction?
• RQ5 How does the proposed multi-scale CNN perform relative to existing methods on the Kaggle dataset?

主要发现

• iEEG channels provide complementary information; excluding channels is not advisable for accurate prediction.
• PCA is not reliable for iEEG data reduction in this seizure prediction context.
• Hand-crafted iEEG features may not generalize well across patients and time.
• The proposed multi-scale CNN approach achieves an average SEN of 87.85% and AUC of 0.84 across patients, outperforming several prior methods.