(B-79) Early detection of liver steatosis using H-scan ultrasound imaging and texture analysis

Introduction:: Poor diet, improper exercise, and other unhealthy lifestyle choices has left much of the general public at risk of compromising their liver’s functionality. As much as 25% of the western society find themselves on the nonalcoholic fatty liver disease (NAFLD) spectrum. However, due to inaccessible or unaffordable testing, many are unable to make educated decisions to improve their quality of life. Deciding the proper protocol for patients on the NAFLD spectrum is currently available after the patient does a liver biopsy or an MRI. Ultrasound affords greater accessibility and affordability to the general population and modern quantifiable parameters are able to characterize key liver features. Our group has used H-scan ultrasound imaging and tissue characterization to detect early steatosis and progression of hepatic steatosis using preclinical animal models. In this current study, we applied our H-scan ultrasound analysis to clinical data and introduce texture analysis for improved characterization of NAFLD as confirmed using magnetic resonance image-derived proton density fat fraction (MRI-PDFF) measures.

Materials and Methods:: Raw ultrasound data from 10 patients was acquired using a commercial ultrasound system equipped with a curvilinear transducer suitable for abdominal scans. An accompanying MRI-PDFF data was acquired for each patient. H-scan ultrasound images and gray level co-occurrence matrix (GLCM) texture analysis of image contrast was calculated independently to determine liver disease severity. H-scan images depict relative ultrasound scatterer size while the GLCM texture analysis measures spatial variance of image intensity values. All ultrasound data processing was performed using custom MATLAB code applied to a region-of-interest encompassing the liver parenchyma.

Results, Conclusions, and Discussions:: H-scan ultrasound imaging revealed a distinction between the clinical normal and stage 1 steatotic livers. However, progression of liver disease correlated with increasing H-scan ultrasound image (R² = 0.88, p = 0.06) and contrast textural features (R² = 0.93, p = 0.03) when compared to patient-matched MRI-PDFF values. Overall, this study supports our hypothesis that H-scan ultrasound images and texture analysis provides a promising noninvasive method for staging liver steatosis in human subjects.

Acknowledgements (Optional): :

References (Optional): :