COMPUTER-ASSISTED AUTOMATED IMAGE RECOGNITION OF CELIAC DISEASE USING CONFOCAL ENDOMICROSCOPY. Enrico Grisan 1 , Hadis Mirzaei 2 , Rupert Leong 2 1 Department of Information Engineering, University of Padova, Italy 2 Faculty of Medicine, The University of New South Wales, Australia ABSTRACT Celiac disease (CD) is an immune-mediated enteropathy triggered by exposure to gluten and similar proteins which affects genetically susceptible persons, increasing their risk of different complications such as malignant lymphomas, small-bowel neoplasia, oropharyngeal tumors, osteoporosis, and bone fractures. In the clinical practice, the diagnosis of CD is made with a diagnostic intestinal biopsy and the concomitant presence of a positive celiac serology. Small bowels mucosa damage due to CD involves various degrees of endoscopically relevant lesions, that are not easily recognized: their overall sensitivity and positive predictive values are poor even when zoom endoscopy is used. Confocal laser endomicroscopy (CLE) has recently revealed to be a useful technique for in vivo virtual histology of small bowels mucosa, but requires skilled and specifically trained operators. We present a computer-based method for the automatic diagnosis of CD-related lesions on the basis of appearance features of confocal images. Comparing the automatic results with the histological gold standard, the proposed method classifies confocal images with accuracy comparable to human observer, suggesting the possibility of real time diagnosis of CD during endoscopy and a non- invasive method to replace biopsy Index Terms— confocal endomicroscopy, computer aided diagnosis, celiac disease, random forests 1. INTRODUCTION Celiac disease (CD) is an immune-mediated enteropathy triggered by exposure to gluten and similar proteins which affects genetically susceptible persons. Celiac disease is one the most frequent enteropathy and most of the celiac patients will remain undiagnosed during their life so it is a hidden epidemic. Exposure to gluten causes damage to the small bowel mucosa. The damage vary from mild, with only an increase in intraepithelial lymphocytes and Crypt Hyperplasia (CH), to severe form, which involves various degrees of endoscopically relevant lesions such as villous Atrophy (VA) [1] (see Fig. 1). VA and CH are important findings that could not be easily recognized during endoscopy; their overall sensitivity and positive predictive values are poor even when zoom endoscopy is used [2]. In today practice, the diagnosis of CD is made with a diagnostic intestinal biopsy and the concomitant presence of a positive celiac serology [3]. Another post-treatment biopsy is sometimes necessary for those patients in whom the first biopsy and serologic test are inconclusive (e.g., seronegative enteropathy) or for patients who remain symptomatic in spite of having strict Gluten Free Diet (GDF) [4]. Introduction of a GFD can significantly improve the symptoms in symptomatic patients, as well as correct the abnormal biochemical measures, therefore improving in patient’s quality of life. Figure 1 Representative images of villi appearance in the normal case (left column), in presence of VA (central column), and CH (right column) Confocal laser endomicroscopy (CLE) is a novel endoscopic method that permits on-site microscopy of the gastrointestinal mucosa after the application of a fluorescent agent, allowing the experienced endoscopists to diagnose VA, CH as well as intraepithelial lymphocytes with high accuracy [5] . The ultimate goal of our study is to provide a system that has the potential to diagnose the CD at the time of endoscopy. To date, few computer aided detection and diagnosis (CADD) systems have been proposed for confocal 978-1-4673-1961-4/14/$31.00 ©2014 IEEE 121