Identifikasi Area Kanker Ovarium pada Citra CT Scan Abdomen Menggunakan Metode Expectation Maximization

Abstract

Ovarian Cancer is a deadly disease, because the patient is too late to be aware of this disease
and come late to treatment. To detect the condition of patient, it’s need examination such as USG with
Doppler, CT scan abdomen or MRI. The examination cannot used to diagnose ovarian cancer, but only to
do operation. Therefore, we need systems to analyze of this condition. One part of the systems is how to
identify area of cancer. In this paper, we use image from ct scan examination result. The method
Expectation Maximization with Gaussian Mixture Model (EM GMM) is used to segmentation of ovarian
cancer areas. The experiment result is EM-GMM method can separate image into some classification
based on pixel feature, even though not so good to distinguish area of cancer and not cancer. It’s seen from
the results of calculation of the percentage of pixels that estimated cancer or not, the value of TP(True Positive) is
45%, while FP(False Positive) is 55%. It caused both of them are same in pixel value. To improve the result,
we need another feature to segmentation, for example is shape feature.
Keywords: CT scan Abdomen, Expectation Maximization, Gaussian Mixture Model, Ovarian Cancer.