An automatic segmentation of endometrial cancer on ultrasound images

Segmentation of endometrial cancer presents a great challenge because these images are distressed by strong speckle noise or noise artifacts. Hence diagnosis of this type of uterus cancer in the premier stage is crucial. The diagnosing procedure is getting ultrasound images of the infected organ and tracing the infected region manually or by applying segmentation algorithms and these segmentation algorithms separate the infected region from the background which easy for the medical practitioners to decide the cancer and steer medication. This paper details two algorithms which is based on level set evolution and an Otsu thresholding method. The US image which is developed from an organ is applied to SRAD filter which is then correlated with various speckle reducing filters such as Lee, Kuan on the basis certain quality metrics measurements like PSNR, MSE, RMSE etc. Similarity metrics namely sensitivity, precision, similarity index have been used for performance evaluation for different methods.