Quantitative Assessment Of Nonalcoholic Fatty Liver Based On Ultrasound RF Signals

Nonalcoholic fatty liver disease(NAFLD)is one of the most common chronic liver diseases worldwide.It is usually at early stage of many liver diseases.Although NAFLD is benign and reversible,there is a high probability to lead to irreversible liver diseases,such as cirrhosis or even liver cancer if it is not diagnosed and treated in time.Therefore,early diagnosis of NAFLD is critical for the patients.Ultrasound imaging is suitable for large-scale screening of NAFLD because it is easy to operate,real-time,no radiation and low cost.However,the resolution of ultrasound imaging is limited,leading to low sensitivity of the diagnosis of NAFLD.Quantitative Ultrasound(QUS)is a new technology developed on the basis of traditional ultrasound imaging.Through processing the ultrasonic signal(Radio-frequency,RF signal),important information related to tissue microstructure could be extracted for disease diagnosis.Single characteristic parameters have different sensitivity to different stages of NAFLD.It is difficult to achieve accurate assessment of each fatty liver staging using a single characteristic parameter.The fusion of multiple characteristic parameters can combine the advantages of each characteristic parameter.This paper aims to establish a quantitative method to diagnose NAFLD through applying the QUS technique.The main work includes:(1)The QUS algorithm based on envelope statistics and frequency domain features is studied separately,including shape parameters in Nakagami statistical distribution,scaling parameters,slope of attenuation coefficient,integral backscattering coefficient,effective scatterer diameter and effective acoustic concentration.The Field II software is used to simulate the ultrasonic echo of the NAFLD condition.The algorithm optimization of various characteristic parameters is realized.(2)Based on the single QUS characteristic feature parameter,the multi-parameter fusion algorithm is developed for the diagnosis of NAFLD,aiming to effectively compensate the insufficiency of single feature parameters.(3)NAFLD auxiliary diagnosis software based on MATLAB environment is developed to realize various functions,including: reading raw data from files in various formats,reconstruction of B-mode ultrasound images,manual selection and preservation of lesion areas,implementation of various algorithms and human-computer interaction interface.(4)The NAFLD animal and clinical data were collected,and the diagnostic performance of the developed QUS algorithm was statistically analyzed.The results show that the multi-parameter QUS model developed in this paper is able to diagnose NAFLD more effectively than the single-parameter model.For the staging of NAFLD animal models,the AUC,sensitivity and specificity of the multi-parameter fusion QUS model were 0.898,0.938 and 0.877,respectively,superior to any single-parameter QUS model.For the staging of clinical NAFLD data,the AUC,sensitivity,and specificity of the multi-parameter fusion QUS model were 0.983,1.0,and 0.9,respectively,superior to any single–parameter QUS model.Conclusively,the multi-parameter fusion QUS method proposed in this paper can make up for the insufficiency of single-parameter QUS method and improve the quantitative staging ability of NAFLD.