基于超声射频信号的大鼠脂肪肝分级诊断研究

Grading Diagnosis of Rats' Fatty Liver Tissue Based on Ultrasonic Radiofrequency Signal

目的:从分析肝脏超声射频信号的角度,探究一种对脂肪肝进行分级定量诊断的新方法。方法:80只Wistar大鼠,通过喂养高脂饲料建立大鼠脂肪肝模型,其中正常肝的大鼠28只,轻度脂肪肝的大鼠21只,中度脂肪肝的18只,重度脂肪肝的13只。然后采集大鼠左右肝的超声射频信号,选取大鼠肝脏部位感兴趣区域的超声射频信号,分析大鼠肝脏超声射频信号的脂肪肝特征信息,提取了射频信号幅度包络值的均值/标准差(MSR)、偏度(SK)、峰度(KU)这三个统计特征量,再利用BP神经网络对大鼠脂肪肝进行分类识别。结果:该方法对大鼠正常肝的识别率达92.5%,轻度脂肪肝的识别率达87.5%,中度脂肪肝的识别率达76.7%,重度脂肪肝的识别率达77.3%。结论:本文研究的方法可对大鼠脂肪肝进行分级定量诊断,且证明了超声射频信号在脂肪肝诊断中是有价值的,为对分级诊断脂肪肝疾病的研究提供了新的方向。

Objective： To explore a new method for the grading diagnosis of fatty liver based on the liver ultrasonic radioffequency （RF） signal. Methods： The 80 fatty liver rat models were established by feeding the rats with a high fat diet. There were 28 rats with normal liver, 21 rats with mild fatty liver, 18 rats with moderate fatty liver, 13 rats with severe fatty liver. The ultrasonic RF signal was acquired from rat＇s left and fight liver, and the RF signal was selected from the region of interest. The fatty characteristics of rat＇s liver RF signal was analyzed, and statistical features were extracted from RF signal envelope. Three features were selected： mean/std （msr）, skewness and kurtosis. Finally, BP neural network was used to classify normal livers and variable degrees of fatty livers. Results： The accuracy rates of classification were 92.5%, 87.5%, 76.7%, 77.3% for normal liver, mild fatty liver, moderate fatty liver and severe fatty liver, respectively. Conclusion： The method of this article is able to achieve the grading diagnosis of rat＇s fatty liver. Ultrasonic RF signal is valuable for the diagnosis of fatty liver. This study provides a new direction for grading diagnosis of fatty liver disease research.