Computer aided diagnosis and grading of stomach adenocarcinoma in microscopic images

Abstract

Stomach cancer is one of the major causes of cancer death worldwide. Almost 90-95% of the stomach (Gastric) cancer is of epithelial in origin. They are called as adenocarcinoma of the stomach. The internal lining of the stomach is composed of glands that are responsible for the secretion of newlinedigestive enzymes. These glands are composed of epithelial cells which secrete newlinethe digestive enzymes into the glandular duct. Malignant tumors arising from newlinethese epithelial cells are known as Adenocarcinoma.Although the latest advances in the imaging technologies have made non-invasive cancer diagnosis a reality, the confirmatory diagnosis is still done newlinebased on the visual examination of the biopsy samples obtained through the newlineendoscopic examination of the patient. On the basis of resemblance of the newlinebiopsies with the normal tissue, the cancerous samples are classified as Well - newlineDifferentiated (Grade I), Moderately - Differentiated (Grade II), Poorly - newlineDifferentiated (Grade III). Diagnosis and Grading are the crucial steps that newlinedetermine the correct choice of treatment offered to the patients to prolong their newlinesurvival. Currently it is performed through the visual examination of newlinepathologists. Their judgment is highly subjected to inter and intra observer newlinevariations. It consumes a lot of time too. In order to overcome these difficulties, newlinethe process of automatic classification of histopathogical stomach biopsies is newlinecarried out. This has paved the way for quantitative assessment of biopsies. newlineIn this thesis, a highly reliable computer aided diagnostic and grading system is proposed for the classification of stomach microscopic images. A comprehensive analysis of the textural and topological variations caused by malignancy in the stomach microscopic images is studied as a novel newlineattempt. As the pathological changes is well known to cause spatial and newline newline