论文部分内容阅读
自动癫痫脑电检测对癫痫的诊断具有重要意义,可以减轻监测长期脑电的工作强度。本文提出和探讨一种基于梯度boosting的长程脑电癫痫检测的新机器学习算法。该算法提取长程脑电的相对波动指数作为特征,采用梯度boosting算法训练分类器来识别发作和正常脑电。最后采用平滑和“collar”技术作为后处理进一步提高检测准确率。利用弗莱堡21位病人的脑电数据对该癫痫检测算法进行评估,实验表明,该算法的平均灵敏度为94.6%,误检率为0.18/h。
Automatic epilepsy EEG detection of epilepsy is important for the diagnosis, can reduce the long-term monitoring of EEG work intensity. This paper presents and explores a new machine learning algorithm based on gradient boosting for long-range EEG detection. The algorithm extracts the relative volatility index of long-term EEG as a feature and uses a gradient boosting algorithm to train the classifier to identify the onset and normal EEG. Finally, smoothing and “collar ” technology are used as post processing to further improve the detection accuracy. The epilepsy detection algorithm was evaluated using EEG data from 21 patients in Freiburg. The experimental results show that the average sensitivity of the algorithm is 94.6% and the false positive rate is 0.18 / h.