遗传程序设计在脑干听觉诱发电位建模中的应用

Brainstem Auditory Evoked Potential Modeling with Genetic Programming

脑干听觉诱发电位(BAEP)对神经学和听力学有着重要价值。利用遗传程序设计(GP)研究BAEP的Ⅴ峰潜伏期,Ⅴ峰和Ⅰ、Ⅲ峰波幅比值与刺激声强之间的函数关系,实现模型的自动获取,以指导临床应用。除基本的遗传操作外,GP方法采用迁移式模型,通过复杂度惩罚控制算法树膨胀。实测数据的建模结果表明,该方法拟合误差较小,性能优于传统GP方法和最小二乘法。这种定量分析将有益于BAEP的临床检测和相关病变诊断。

Brainstem auditory evoked potential (BAEP) is of important significance in neurology and audiology. Function relations between V peak latency and stimulus intensity, between V/I, V/III amplitude ratio and stimulus intensity in BAEP are investigated with genetic programming (GP). Models are obtained automatically to guide the clinic application. Sub-population migration and tree inflation controlled by the complexity penalty are applied in GP besides basic operators. Modeling results show that the fitting error is smaller, and the performance exceeds traditional GP and least-squares algorithm. The quantitative analysis related with the curve fitting is helpful to the clinic detection of BAEP and disease diagnosis.