Observing and classifying the indirect immunofluorescence patterns on HEp-2 cells can help in detecting Anti-Nuclear-Antibodies. A computer algorithm to perform this function can lead to a more standardized, faster and accurate diagnosis of auto-immune diseases such as systemic lupus erythematosus, Sjogren's syndrome, and rheumatoid arthritis. In this paper, HEp-2 staining patterns are classified using segmentation based fractal texture features. The images used for this experimentation are obtained from a publicly available database. The features extracted from a cell image is used to classify it into homogenous, fine speckled, coarse speckled, centromere and nucleolus. The cell images are segmented using the ground truth mask provided in the database. Adaptive histogram equalization is applied to the segmented images for contrast enhancement. Three features namely mean intensity, area and Hausdorff fractal dimension of the border are extracted for 8 different Otsu threshold levels. Finally, the 24 features thus extracted are fed to a support vector machine with Gaussian radial basis function kernel. It is observed that the overall accuracy of classification is 65.17%. The accuracy is greatly dependent on scaling and distribution of the features given to SVM. It appears that the segmentation based fractal texture features and SVM could help to build a robust automated diagnosis tool for auto-immune diseases.