Executive Summary : | Fatty liver or nonalcoholic steatohepatitis (NASH) is being recognized as a silent epidemic of the world. Fatty liver prevalence is estimated to be around 9-32 % in the Indian population, with a higher incidence rate amongst obese and diabetic patients. Diagnosis of fatty liver in its initial stage is difficult and can turn costly. Radiological imaging based on MRI and CT is unaffordable and inaccessible for a large population of India. On the other hand, ultrasound based methods are cheaper, easily portable, and can be made widely accessible to provide low-cost fatty liver diagnosis. Changes in viscoelastic properties of liver tissue may often be symptomatic of dysfunction that can be correlated to tissue pathology. Thus, viscoelastic parameters of liver tissue can act as biomarkers to provide information about steatosis grade in the diseased liver. Shear wave elastography can be developed into a suitable diagnosis technique for this purpose. Shear waves can be generated inside a soft tissue either by sending a focused ultrasound beam with a conventional ultrasound probe or through external vibrations. This causes a few micrometers displacement within the tissue that propagates as a transient shear wave. The shear waves can be tracked with the same ultrasound probe. In general, the aim is to indirectly measure the viscoelasticity of the tissue by monitoring the propagation of shear waves inside it. Shear waves are known to propagate faster in stiffer media and slower in softer media. Development of advanced imaging models for ultrasound elastography are proposed here for improving fatty liver or NASH diagnosis. Generation of the steady and potent shear wave inside thick tissues and deeper locations is a challenge. In addition, the tissue structure and boundaries cause multiple reflections that create unwanted noise levels and artifacts in the data which needs to be pre-processed. Image reconstruction models relying on a low shear wave (SW) frequency content can be useful in imaging deeper issue locations as lower frequency attenuates less. The development of multi-directional noise filtering models is required to address the measurement noise challenge. Deep learning models will be developed to perform intelligent image reconstruction in ultrasound elastography. These models will be developed to improve elastography image features to better assist radiologists. Biomarkers based on tissue viscoelasticity will improve fatty liver diagnosis and patient monitoring. A major impact of this work will provide a non-invasive and cost-effective technique for the detection of liver steatosis and identification of patients with the advanced forms: steatohepatitis, fibrosis, and cirrhosis induced by nonalcoholic steatohepatitis (NASH). Moreover, these techniques could also be utilized for other soft tissue related diseases such as deep vein thrombosis (DVT), breast cancer, etc. |