Image processing methods for computer-aided interpretation of microscopic images

Abstract

Image processing algorithms for automated analysis of microscopic images have become increasingly popular in the last decade with the remarkable growth in computational power. The advent of high-throughput scanning devices allows for computer-assisted evaluation of microscopic images, resulting in a quick and unbiased image interpretation that will facilitate the clinical decision-making process. In this thesis, new methods are proposed to provide solution to two image analysis problems in biology and histopathology. The first problem is the classification of human carcinoma cell line images. Cancer cell lines are widely used for research purposes in laboratories all over the world. In molecular biology studies, researchers deal with a large number of specimens whose identity have to be checked at various points in time. A novel computerized method is presented for cancer cell line image classification. Microscopic images containing irregular carcinoma cell patterns are represented by subwindows which correspond to foreground pixels. For each subwindow, a covariance descriptor utilizing the dual-tree complex wavelet transform (DTCWT) coefficients as pixel features is computed. A Support Vector Machine (SVM) classifier with radial basis function (RBF) kernel is employed for final classification. For 14 different classes, we achieve an overall accuracy of 98%, which outperforms the classical covariance based methods. Histopathological image analysis problem is related to the grading of follicular lymphoma (FL) disease. FL is one of the commonly encountered cancer types in the lymph system. FL grading is based on histological examination of hematoxilin and eosin (H&E) stained tissue sections by pathologists who make clinical decisions by manually counting the malignant centroblast (CB) cells. This grading method is subject to substantial inter- and intra-reader variability and sampling bias. A computer-assisted method is presented for detection of CB cells in H&Estained FL tissue samples. The proposed algorithm takes advantage of the scalespace representation of FL images to detect blob-like cell regions which reside in the scale-space extrema of the difference-of-Gaussian images. Multi-stage false positive elimination strategy is employed with some statistical region properties and textural features such as gray-level co-occurrence matrix (GLCM), gray-level run-length matrix (GLRLM) and Scale Invariant Feature Transform (SIFT). The algorithm is evaluated on 30 images and 90% CB detection accuracy is obtained, which outperforms the average accuracy of expert hematopathologists.