Urinary bladder tumor grade diagnosis using on-line trained neural networks

Tasoulis, D. K. and Spyridonos, P. and Pavlidis, N. G. and Cavouras, D. and Ravazoula, P. and Nikiforidis, G. and Vrahatis, M. N. (2003) Urinary bladder tumor grade diagnosis using on-line trained neural networks. In: Knowledge-Based Intelligent Information and Engineering Systems : 7th INternational Conference, KES 2003, Oxford, UK, September 2003. Proceedings, Part I. Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science), 2773 P . Springer, GBR, pp. 199-206. ISBN 9783540408031

Abstract

This paper extends the line of research that considers the application of Artificial Neural Networks (ANNs) as an automated system, for the assignment of tumors grade. One hundred twenty nine cases were classified according to the WHO grading system by experienced pathologists in three classes: Grade I, Grade II and Grade III. 36 morphological and textural, cell nuclei features represented each case. These features were used as an input to the ANN classifier, which was trained using a novel stochastic training algorithm, namely, the Adaptive Stochastic On-Line method. The resulting automated classification system achieved classification accuracy of 90%, 94.9% and 97.3% for tumors of Grade I, II and III respectively.