Using Short Time Fourier Transform To Ultrasound Signals For Fatty Liver Detection

Fatty liver disease (FLD) occurs in 15% of the general population, and it has a prevalence of 75% among obese persons. FLD is the cause of chronic liver injury, which will lead to fibrosis and cirrhosis. So it is important to diagnose the FLD for most patients. Liver biopsy is the clinical method for detecting and assessing FLD. However, the biopsy program is invasive, and it may cause some different complications, such as internal hemorrhage. Furthermore, the sample volume is limited so it can not reflect the status of the whole liver accurately. Particularly, FLD is a chronic process of accumulation of fats, therefore patients who do not have the obvious clinical symptoms usually have no intention to go through a liver biopsy. For these reasons, non-invasive techniques should be taken into consideration to diagnose FLD. Magnetic resonance imaging (MRI), computed tomography (CT) and ultrasonography (US) are the most used noninvasive techniques. Among these, US is the most important tool for evaluating FLD because it is inexpensive and real-time. The Ultrasound B-mode image shows brighter gray levels which can give physicians basic information to judge the status of FLD. It means that the ultrasound imaging is highly subjective and depends on the user experiences. Therefore some objective methods of analysis are needed. Generally speaking, some objective approaches, such as texture analysis and quantitative ultrasound (QUS) techniques, have been widely applied in FLD detection. In QUS based imaging many parameters are estimated by using frequency or time domain based approaches, such as sound speed, backscatter coefficient and attenuation coefficient. Attenuation coefficient is one of the parameters which can help us estimate FLD, because fatty infiltration increases the acoustic attenuation. Note that attenuation results in the downshift in ultrasound frequency. If using a general Fourier transform to analyze the change in frequency, the temporal resolution of spectrum may be limited. In this study, we tried to use short-time Fourier transform (STFT) as a method to evaluate the change in signal frequency to classify the stage of FLD by ultrasound. STFT may be a suitable choice as it can be used in vivo and real-time. To the best of our knowledge, no any literature about using STFT for fatty liver detection was found.