Abstract—A wide range of different image modalities for medical imaging are available now-a-days which provide view of internal structures of the human body such as brain, kidney, liver etc. Among these medical image modalities, Ultrasound (US), Magnetic Resonance Imaging (MRI), Computed Tomography (CT), Positron Emission Tomography (PET) and Positron Emission Tomography combined Computed Tomography (PET/CT) imaging has gained importance in research areas. The key problems within medical image analysis techniques are segmentation and shape feature extraction that will be referred to in this paper. Segmentation of grayscale medical images can be difficult since the intensity values between healthy tissue and tumor may be very close. PET/CT provides more accurate measurements of tumor size than is possible with visual assessment alone. In this paper, segmentation method for the detection of liver tumor in PET/CT scans is proposed. The images are denoised using median filter and binary tree quantization clustering algorithm is used for segmentation. Finally ROI selection and shape feature extraction is performed on the selected cluster to quantify the size of the tumor and to check the accuracy of our method with original image and K-means Clustering method.

Index Terms—denoising, liver segmentation, binary tree quantization algorithm, k-means clustering algorithm, shape feature extraction

Cite: Neha Bangar and Akash Deep, "Tumor Detection by Color Segmentation of PET/CT Liver Images," Journal of Emerging Technologies in Web Intelligence, Vol. 6, No. 1, pp. 148-156, February 2014. doi:10.4304/jetwi.6.1.148-156