Data mining framework for fatty liver disease classification in ultrasound : a hybrid feature

Fatty Liver Disease (FLD) is an increasing prevalent disease that can be reversed if detected early. Ultrasound is the safest and ubiquitous method for identifying FLD. Since expert sonographers are required to accurately interpret the liver ultrasound images, lack of the same will result in inter-observer variability. For more objective interpretation, high accuracy, and quick second opinions, Computer Aided Diagnostic (CAD) techniques may be exploited. In this paper, we present a CAD technique (a class of Symtosis, by Global Biomedical Technologies Inc., CA) that uses significant features such as texture, wavelet transform and higher order spectra in various supervised learning based classifiers in order to determine parameters that classify normal and FLD-affected abnormal livers. On evaluating the proposed technique using 20 abnormal and 15 normal liver ultrasound images, we were able to achieve a high classification accuracy of 93.3% using several classifiers. This high accuracy added to the completely automated classification procedure makes our proposed technique highly suitable for clinical deployment and use, once the technique is verified on a larger database of liver images. We have also proposed a novel SteatosisClassificationIndex (SCI) that uses a combination of the significant features to output a single number which can be used to determine the class of the tested image. This index can also be easily incorporated into a clinical setting at no extra cost and it can provide additional diagnostic confidence to the physician.