Abstract：Objective We used a smoothed noisy intensity profile by a sigmoid function and employ it to discover the true location
of CT/MR tumor boundary more accurately. Methods A novel combination of the support vector machine, watershed, and scattered
data approximation algorithms were employed to initially segment a tumor. Small and large abnormalities were treated distinctly.
Next, the proposed sigmoid edge model was fitted to the normal profile of the border. The estimated parameters of the model were
then utilized to find true boundary of a tissue. The quantitative metrics were evaluated by liver segmentation challenge proposed
by Medical Image Computing and Computer Assisted Intervention. Results We extensively evaluated our method using synthetic
images (contaminated with varying levels of noise) and clinical CT/MR data. Based on the sensitivity analysis results, we decided
to set the threshold for data approximation, number of sectors and dgap as 15, 12 and 4, respectively. Visually and quantitatively
experimental results indicated that VOE and RVD of the proposed method were 28.21% and 19.20% in the first team and 7.62% and
13.45% in the second team, which were superior to the existing methods. Conclusion For different size and any types of tumors, the
proposed method can obtain more efficient and accurate segmentation results. It can also provide better robustness, superiority, and
pervasiveness in the noise environment and clinical applications.
