Segmentation of Liver Metastases Using a Level Set Method with Spiral-Scanning Technique and Supervised Fuzzy Pixel Classification Dirk Smeets, Bert Stijnen, Dirk Loeckx, Bart De Dobbelaer and Paul Suetens Katholieke Universiteit Leuven ESAT - PSI dirk.smeets@gmail.com Abstract. In this paper a specific method is presented to facilitate the semi-automatic segmentation of liver metastases in CT images. Accurate and reliable segmentation of tumors is e.g. essential for the follow-up of cancer treatment. The core of the algorithm is a level set function. The initialization is provided by a spiral-scanning technique based on dy- namic programming. The level set evolves according to a speed image that is the result of a statistical pixel classification algorithm with super- vised learning. This method is tested on CT images of the abdomen and compared with manual delineations of liver tumors. The results show that the accuracy of the method does not reach that of the manual seg- mentation. The average overlap error is 34.6% and the average volume difference is 17.8%. The average, the rms and the maximum surface are respectively 2.0, 2.7 and 10.1 mm. 1 Introduction Segmentation is an image processing operation to distinguish an anatomical structure from the surrounding tissue. Tumor segmentation is an important issue for cancer follow-up, where the oncologist is interested to evaluate the change in size of the tumors. Early response prediction allows the oncologist to adapt the therapy, which can lead to a higher survival rate [1]. Measuring the response of a treatment can be done by uni-, bi- or tridimensional criteria. Clinical research [1– 3] indicates that volume measurements (3D) give the best reflection of the tumor response. Volume measurements require the segmentation of the tumors, which is very time consuming when done manually. Moreover, the volume of manual delineations is subject to intra- and interobserver variability, which is estimated at about 8% for liver tumors [4]. Therefore, automatic or semi-automatic tumor delineation algorithms are required. This article focuses on the segmentation of liver tumors in contrast-enhanced CT images. Because tumors generally have different shapes and intensities, the segmentation is not straightforward. The gray values of a tumor depend on the delay between the contrast injection and the image acquisition, the contrast dose and the patient physiology. In general, liver tumors have a more or less round