Principal component analysis based algorithms for ECG recordings

记录在心电中的P-QRS-T时间波隐藏了对检测心血管疾病至关重要的信息，人们致力于开发基于心电的方法来区分规则心跳和不规则心跳，并对心律失常进行检测和分类。这一领域的研究和开发已经**保持活跃了几十年，因为在检测和分类能力方面的任何改进**都是增强心脏健康监测系统的迫切需要。**心脏通讯解决方案公司一直在医疗诊断行业，作为一家FDA批准、ISO认证和**加拿大卫生部/CE批准的医疗软件和设备的开发、销售和营销公司。**该公司的专长是基于计算机的心电管理和报告软件的软件工程**。该公司目前正在寻求增强和扩展其心电分析软件，并**特别感兴趣的是开发用于自动分析来自各种**心电设备的心电记录的算法，这些设备具有不同质量和采样率的不同数据量。一套起源于统计学中的多元分析的技术，称为主成分分析(PCA)，已被该公司选为心电分析的基础工具。这项建议将解决**公司在这一领域的开发和实践中产生的主要问题，包括：(I)使用**MIT-BIH心律失常数据库训练的PCA子空间的普遍性；(Ii)处理不同采样率的心电记录的技术；(Iii)**确定QRS波群的最佳聚类方法；以及(Iv)PCA特征空间的某些部分(就形状和大小而言)与已知QRS形态之间是否存在内在联系。**预期的结果将大大增强心脏通信公司的解决方案组合，并向消费者提供自动化和**准确的心电分析。

The P-QRS-T time waves recorded in electrocardiogram (ECG) conceal information vital for detecting the**cardiovascular disease, and much effort has been made to develop ECG-based methods that distinguish regular**from irregular heartbeats, and detect and classify heart arrhythmia. Research and development in this field have**stayed active for decades as any improvement in accuracy, speed, and robustness in detection and classification**capabilities is highly desirable for enhancing cardiac health monitoring systems.**CardioComm Solutions Inc. has been in medical diagnostic industry as an FDA cleared, ISO certified, and**Health Canada/CE approved company for development, sales, and marketing of medical software and devices.**The company's specialization is in the software engineering of computer based ECG management and reporting**software. The company is currently looking to enhance and extend its software for ECG analysis, and is**especially interested in developing algorithms for automatic analysis of ECG recordings coming from a variety**of ECG devices with different data sizes of varying quality and sampling rates. A suite of techniques originated**from multivariate analysis in statistics, known as principal component analysis (PCA), has been selected by the**company as a foundational tool for ECG analysis. This proposal will solve the major issues arising from the**company's development and practice in this area include (i) universality of the PCA subspaces trained using**MIT-BIH arrhythmia database; (ii) techniques to handle ECG recordings with different sampling rates; (iii)**identification of optimal methods for clustering QRS complexes; and (iv) existence of intrinsic connections, if**any, between certain parts (in terms shape and size) of PCA feature space and known QRS morphologies. The**expected outcome will significantly enhance CardioComm's solution portfolios and provide automated and**accurate ECG analysis to consumers.