سمپوزیوم منطقه ای نوآوری در علم و فناوری

English

صفحه اصلی / دومین سمپوزیوم منطقه ای نوآوری در علم و فناوری

Classification of Non-Alcoholic Fatty Liver Disease Using Ultrasound Imaging and Deep Texture Representation

نویسندگان :

ALI FAROOGHI¹

1- دانشگاه آزاد اسلامی واحد بیرجند

کلمات کلیدی :

NAFLD،Ultrasound،Deep Learning،Statistical Analysis،Convolutional Neural Networks

چکیده :

Abstract—Non-Alcoholic Fatty Liver Disease (NAFLD) is one of the most prevalent chronic liver diseases globally, driven by increasing obesity rates and unhealthy lifestyles. Early diagnosis of NAFLD is crucial for preventing its progression to more severe stages, such as steatohepatitis and cirrhosis. Ultrasound imaging is widely employed for screening and monitoring NAFLD due to its non-invasive nature and accessibility. However, manual interpretation of ultrasound images can be subjective and prone to inaccuracies. This study introduces an automated classification method for NAFLD based on deep texture representation using Convolutional Neural Networks (CNN) and statistical analysis. Ultrasound images from 220 patients (120 with NAFLD, 100 healthy) were processed using MATLAB, and the performance of the model was evaluated through statistical analysis using SPSS. The proposed model achieved an accuracy of 94.3%, sensitivity of 91.8%, and specificity of 96.1%. The impact of different texture parameters on the model’s performance was analyzed, and a comparison with other machine learning approaches was provided.