计算机医用信号处理与分析——基于小波变换的医学瞬变信号去噪新方法

A NOVEL METHOD FOR MEDICAL SIGNAL DENOISING BASED ON WAVELET TRANSFORM

本文提出一种用于计算机辅助诊断前处理的医学信号去噪新方法。方法：１．对信号进行ＤＷＴ。２．在ｓ＝２１上，利用邻近尺度ＷＴ系数的空间相关性进行选择性去噪。３．在ｓ≥２２上，运用阈值法（Ｓｏｆｔ，ＨａｒｄＴｈｒｅｓｈｏｌｄｉｎｇ）去噪。４．信号重构。结果：分别采用本文方法及传统方法对突变信号进行去噪与结果比较。结论：本文方法在获得信噪比增益的同时，能够保持对信号突变信息的良好分辨且处理速度快、便于在微机上实现。

Purposes: This paper presents a robust method of medical transient signals de-noisingfor computer-aided diagnosis system based on wavelet transform(WT). Methods: 1.Waveletdecomposition of signal at 5 scales. 2. At s=21, Spatially selective noise filtration technique wasused in WT domain to reduce noise. 3. At s≥22,wavelet shrinkage techniques( hard thresholdingand soft thresholding ) were applied to further reduce noise. 4. Signal reconstruction. Results: Theperformance of the technique was compared to that of the traditional methods and found it to besuperior. Conclutions: The SNR can be substantially increased by this method with fine detection of transients, small features retention and high efficiency on conventional computers.