| Summary: | 碩士 === 國立成功大學 === 數學系應用數學碩博士班 === 102 === In this study, we propose a statistical model for assessing liver fibrosis by norm training. We assume that the surface of the healthy liver is uniform, as well as the absorption, reflection and scattering behaviors of ultrasound. In addition, we did not assume that the echo intensity obeys some fixed kind of distributions. In practice, the intensity of ultrasound decays as the depth increases. However, the intensity should be uniform for the later analysis. It forces us to construct a suitable compensation for different depths. The compensation is done by the least-square method derived from fitting the average amplitude of the radio frequency data. We use the differences between the cumulative distribution functions for the comparison and use the radio frequency data from six health livers to establish the norm to examine the distinction of the different stage of fibrosis.
By comparing the difference from the norm, the cumulative distribution function provides a quantitative base to represent the difference of normal and lesion tissue. Different degrees of liver fibrosis can be revealed by the assessment.
|
|---|
