Quantification of Hepatic Fat Level Using Gray Level Co-Occurrence Matrix (GLCM) and Extreme Learning Machine (ELM) Methods in Ultrasonography Image (USG)

Abstract

Ultrasonography (USG) is one of the alternative modalities used to avoid biopsies in evaluating Non-alcoholic Fatty Liver Disease (NAFLD). But having weaknesses is subjective and the results of the examination are very dependent on the ability, expertise, skills and experience of the operator. The application program of the GLCM and ELM methods provides a second opinion and increases the operator's subjectivity in interpreting the results of the ultrasound examination. Analyze the ability of the GLCM and ELM methods in the compatibility of fatty liver levels in ultrasound images, with indicators of sensitivity values, specificity and accuracy of the application program.
This study used a retrospective sample conducted at Semarang Medical Center (SMC) in Telogorejo Hospital, Semarang. The research test samples were 108 then carried out a diagnostic test to determine the sensitivity, specificity and accuracy of the application program.
Based on the diagnostic test obtained a high sensitivity value. At normal there are 93%, grade 1 is 74%, grade 2 is 78%, and grade 3 is 85%. Specificity of normal fatty liver is 100%, grade 1 is 98%, grade 2 is 86%, grade 3 is 93%. Normal accuracy is 98%, grade 1 is 92%, grade 2 is 84%, grade 3 is 91%..
The GLCM and ELM methods are good at detecting fatty liver levels in ultrasound images.