HEp-2 Cells classification via fusion of morphological and textural features

Abstract

Autoimmune diseases are proven to be connected with the occurrence of autoantibodies in patient serum. Antinuclear autoantibodies (ANAs) identification can be accomplished in a laboratory using indirect immunofluorescence (IIF) imaging. ANAs are characterized by specific "visual" patterns on a humane epithelial cell line (HEp-2). The identification stage is usually done by trained and highly qualified physicians through visual inspection of slides using a fluorescence microscope. The presence of subjectivity in the identification process, the interobserver variability, the increasing demand of highly specialized personnel, suggest that a realization of an automatic classification system is of great significance for the field of autoimmune diseases diagnosis. Moreover CAD systems can be used in a collaborative scheme in order to augment the physicians' capabilities. In this paper a system for automatic classification of staining patterns on single-cell fluorescence images is proposed. Our method utilizes morphological features extracted from a set of binary images derived via multi-level thresholding of fluorescence images. Furthermore, a modified version of Uniform Local Binary Patterns descriptor is incorporated in order to capture local textural information. The classification is performed using a non-linear SVM Classifier. The proposed method is evaluated using a publicly available dataset, recently released for the purposes of HEP-2 Cells classification competition at ICPR 2012, achieving up to 95.9% overall classification accuracy.

Automatic Classification of Human Epithelial Type 2 Cell Indirect Immunofluorescence Images using Cell Pyramid Matching

“…textural) for each class Multi Expert System Cordelli et al [7] Image statistics; textural; morphological AdaBoost Strandmark et al [35] Morphological; image statistics; textural Random Forest Ali et al [2] Biological-Inspired Descriptor Boosted k-NN Classifier Theodorakopoulos et al [36] Morphological and texture features Kernel SVM (KSVM) Thibault et al [37] Morphological and texture features Linear Regression, Random Forest Ghosh et al [13] Histograms of Oriented Gradients, SVM image statistics and textural Li et al [19] Textural and image statistics SVM Di Cataldo et al [5] GLCM and DCT features SVM Snell et al [33] Texture and shape Multistage classifier Ersoy et al [9] Local shape measures, gradient and textural ShareBoost Wiliem et al [41] Bag of visual words with dual-region structure Nearest Convex Hull Classifier (NCH) and apply an automated feature selection process [15]. Another approach uses Multi Expert Systems to allow the use of a specifically tailored feature set and classifier for each HEp-2 cell pattern class [34].…”

HEp-2 staining pattern recognition at cell and specimen levels: Datasets, algorithms and results

“…Successively, a further noise reduction step is performed: the image is divided in patches that are encoded through sparse representation using a dictionary learned on low-noise images. Then the images are reconstructed, normalized again and classified using a multi-class SVM with linear kernel fed with a vector including both morphological and textural features as in Theodorakopoulos et al (2012). In order to extract the former, the image is thresholded using 14 equally spaced values.…”

HEp-2 fluorescence pattern classification

“…Another way of comparing these textures, used by a number of contest participants [15,11,12], is granulometry or morphological measurements of image slices at different thresholds. Similarly to [15], we consider 7 thresholds equally spaced between the extremes of intensity within each image, and compute 3 parameters from the connected objects produced at each threshold:…”

“…• average circularity of all the objects Again following [15], we filter out objects below a certain size as noise. The resulting descriptor has 7 * 3 = 21 features.…”

“…Similarly to [15], we consider 7 thresholds equally spaced between the extremes of intensity within each image, and compute 3 parameters from the connected objects produced at each threshold:…”