Yang Danni, Lin Wentao, Liu Minghui, Zhou Yuanfeng, Wang Yalin
School of Materials Science and Engineering, Lanzhou University of Technology, Lanzhou 730050, People's Republic of China.
State Key Laboratory of Advanced Processing and Recycling of Non-ferrous Metals, Lanzhou University of Technology, Lanzhou 730050, People's Republic of China.
J Neural Eng. 2025 Jan 17;22(1). doi: 10.1088/1741-2552/ada0e7.
Measuring causal brain network from neurophysiological signals has recently attracted much attention in the field of neuroinformatics. Traditional data-driven algorithms are computationally time-consuming and unstable due to parameter settings.To resolve these limits, we proposed a novel parameter-free technique, called 'non-parametric full cross mapping (NFCM)'. The NFCM adapts current convergent cross-mapping concept, and makes two improvements: (1) an improved phase-space reconstruction with constant embedding parameters and (2) cross-mapping estimate of all embedding vectors on manifolds following simplex projection.Numerical experiments verify that our NFCM has the highest quantization stability even when perturbed by system noise, and its coefficient of variation is almost lower than that of the six baseline methods. The developed NFCM is finally used in stereoelectroencephalogram analysis of drug-resistant epilepsy in children (DREC). A total of 36 seizures, comprising 18 surgical successes and 18 failures, were included to explore the brain network dynamics. The average causal coupling in epileptogenic zones of successful surgery (0.81 ± 0.04) is significantly higher than that in non-epileptogenic zones (0.40 ± 0.03) withP<0.001via Mann-Whitney-U-test. While there is no significant difference among the 18 failed surgeries.The causal brain network measured by our NFCM is confirmed as a credible biomarker for localizing epileptogenic zones in DREC. These findings promise to advance precision medicine for DREC.
从神经生理信号中测量因果脑网络最近在神经信息学领域引起了广泛关注。传统的数据驱动算法由于参数设置的原因，计算耗时且不稳定。为了解决这些局限性，我们提出了一种新颖的无参数技术，称为“非参数全交叉映射（NFCM）”。NFCM采用了当前的收敛交叉映射概念，并进行了两项改进：（1）使用恒定嵌入参数的改进相空间重构；（2）在单纯形投影后的流形上对所有嵌入向量进行交叉映射估计。数值实验验证了我们的NFCM即使在受到系统噪声干扰时也具有最高的量化稳定性，并且其变异系数几乎低于六种基线方法。最终，所开发的NFCM被用于儿童耐药性癫痫（DREC）的立体脑电图分析。共纳入36次癫痫发作，其中18次手术成功，18次失败，以探索脑网络动力学。通过曼-惠特尼-U检验，成功手术的致痫区平均因果耦合（0.81±0.04）显著高于非致痫区（0.40±0.03），P<0.001。而18次失败手术之间没有显著差异。我们的NFCM测量的因果脑网络被确认为DREC中致痫区定位的可靠生物标志物。这些发现有望推动DREC的精准医学发展。
