基于分数阶Hilbert变换二维纹理特征的罗音检测算法

Algorithm of Crackle Detection Based on Two-Dimensional Texture Features by Fractional Hilbert Transform

为了自动化地检测肺音中间断性异常呼吸音——罗音,基于新兴的分数阶Hilbert变换信号分析手段,提出一种新的检测算法。在不同阶数下,将分数阶Hilbert变换作用于肺音信号,由此可生成一个与该信号对应的二维纹理图像,纹理图像中呈现的明暗相间条纹组合的特征与罗音信号特征相对应,据此可实现罗音检测。分别进行两组检测实验,一组采用仿真的罗音叠加于实际肺音上进行罗音检测,100%的罗音被检测出,另一组采用美国胸科学会教学数据库包含真实罗音的肺音信号进行检测,敏感性指标达到97%,从而验证基于分数阶Hilbert变换的方法检测罗音的有效性。

In order to detect an important kind of abnormal and discontinuous respiratory sounds crackles in lung sounds, a new detection algorithm has been proposed based on an emerging theory of fractional Hilbert transform. By applying fractional Hilbert transform to lung sound signals, a two-dimension image with texture feature can be generated. The features of combination of dark and bright interlaced strips in texture images were corresponding to features of crackle signals in lung sounds, which can be employed to detect crackles. Two groups of detection experiments were carried out. First, real normal lung sounds signals with simulated crackles embedded were tested, and 100% crackles were detected. Then, lung sounds of patients with pulmonary disease were tested, and the sensitivity of detection accuracy reached to 97%. Thus, the effectiveness of crackle detection based on fractional Hilbert transform has been validated.