13 Dounias

In this study the performance of various intelligent methodologies is compared in the task of pap-smear diagnosis. The selected intelligent methodologies are briefly described and explained, and then, the acquired results are presented and discussed for their comprehensibility and usefulness to medical staff, either for fault diagnosis tasks, or for the construction of automated computer-assisted classification of smears. The intelligent methodologies used for the construction of pap-smear classifiers, are different clustering approaches, feature selection, neuro-fuzzy systems, inductive machine learning, genetic programming, and second order neural networks. Acquired results reveal the power of most intelligent techniques to obtain high quality solutions in this difficult problem of medical diagnosis. Some of the methods obtain almost perfect diagnostic accuracy in test data, but the outcome lacks comprehensibility. On the other hand, results scoring high in terms of comprehensibility are acquired from some methods, but with the drawback of achieving lower diagnostic accuracy. The experimental data used in this study were collected at a previous stage, for the purpose of combining intelligent diagnostic methodologies with other existing computer imaging technologies towards the construction of an automated smear cell classification device. Introduction This report is the result of extensive collaborative work between engineers and doctors of different fields of expertise originating from different EU countries. The implementation of an efficient computer-assisted methodology for automated classification of cell images is important because manual screening is a tedious and error prone task. Various intelligent methods for data analysis and knowledge extraction are briefly explained and their performance is compared in terms of effectiveness, accuracy and comprehensibility, in the medical domain of pap-smear diagnosis. Extensive details for some of the intelligent methods applied to pap-smear diagnosis, together with related comparisons of performance and findings, can be found in refs. 1-3. According to Meisels and Morin (4), cervix is the anatomical region between the body of the uterus and the vagina and is covered by epithelium. In the vaginal part of the cervix there is stratified squamous epithelium, in the endocervical part of the cervix there is simple columnar epithelium and between them there is the transitional zone. Histologically, the cells of the squamous epithelium are arranged in four layers: the basal, the parabasal, the intermediate and the superficial layer. The columnar epithelium consists only of one layer. Using a small brush, a specimen is taken from the cervix and transferred on to a slide. The smear is stained using the Papanikolaou method thus making it possible to see characteristics of cells on the microscope and classify them according to the homonymous classification. The purpose of Papanikolaou classification (3,4) is to diagnose pre-malignant changes before they progress to invasive carcinoma. In normal columnar epithelial cells, the nucleus is located at the bottom of the cytoplasm. When viewed from the top, the nucleus seems larger. When viewed from the side, the cytoplasm seems larger. Cells of the basal and parabasal layer have small nuclei and cytoplasm. Cells of intermediate and superficial layer have small nuclei and larger cytoplasm. Squamous dysplastic cells generally, have larger and darker nuclei and tend to cling together in clusters. ONCOLOGY REPORTS 15: 1001-1006, 2006 Automated identification of cancerous smears using various competitive intelligent techniques G. DOUNIAS1, B. BJERREGAARD2, J. JANTZEN3, A.TSAKONAS4, N. AMPAZIS1, G. PANAGI5 and E. PANOURGIAS6 1Department of Financial and Management Engineering, University of the Aegean, 31 Fostini Str., 82100 Chios, Greece; 2Herlev University Hospital, DK-2730 Herlev; 3Oersted-DTU Automation, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark; 4Artificial Intelligence and Information Analysis Laboratory, Department of Informatics, Aristotle University of Thessaloniki, 54124 Thessaloniki, 5Department of Radiology, General Hospital of Chios ‘Skilitsion’, Chios; 6Department of Radiology, Euroclinic Hospital, 9 Athanasiadou Str., 11521 Athens, Greece Received September 6, 2005; Accepted September 28, 2005 _________________________________________ Correspondence to: Dr Georgios Dounias, Department of Financial and Management Engineering, University of the Aegean, 31 Fostini Str., 82100 Chios, Greece E-mail: [email protected]