Abreu-Lima C, de Sá J P
Serviço de Medicina II, Faculdade de Medicina, Porto.
Rev Port Cardiol. 1998 May;17(5):415-28.
The morphological diagnosis of ECGs is a pattern recognition procedure. The way the clinician does this is not clearly elucidated. Nevertheless, several models aimed at achieving identical results by automatic means are empleyed. While in the doctor's case this is not exactly so, the computer task for ECG interpretation comprises two distinct and sequential phases: feature extraction and classification. A set of signal measurements containing information for the characterization of the waveform is first obtained. These waveform descriptors are then used to allocate the ECG to one or more diagnostic classes in the classification phase. The classifier can embody rules-of-thumb used by the clinician to decide between conflicting ECG diagnosis and formal or fuzzy logic as a reasoning tool (heuristic classifiers). On the other hand, it can use complex and even abstract signal features as waveform descriptors and different discriminant function models for class allocation (statistical classifiers). More recently, artificial neural network techniques have also been used for signal classification. The authors review feature selection techniques and classification strategies, problems and methods of performance evaluation and results obtained by different classification approaches. A brief discussion of the relative merits of the two main types of ECG classifiers, logical and statistical, is included.
心电图的形态学诊断是一个模式识别过程。临床医生进行此项诊断的方式尚未完全阐明。然而，人们采用了几种旨在通过自动手段获得相同结果的模型。虽然医生的情况并非完全如此，但心电图解读的计算机任务包括两个不同且相继的阶段：特征提取和分类。首先要获得一组包含用于表征波形信息的信号测量值。然后在分类阶段，这些波形描述符被用于将心电图分配到一个或多个诊断类别。分类器可以体现临床医生用于在相互矛盾的心电图诊断之间做出决定的经验法则以及作为推理工具的形式逻辑或模糊逻辑（启发式分类器）。另一方面，它可以使用复杂甚至抽象的信号特征作为波形描述符以及用于类别分配的不同判别函数模型（统计分类器）。最近，人工神经网络技术也已用于信号分类。作者回顾了特征选择技术和分类策略、性能评估的问题和方法以及不同分类方法所获得的结果。文中还简要讨论了两种主要类型的心电图分类器，即逻辑分类器和统计分类器的相对优点。
