Algoritmi di supporto alla diagnosi istopatologica delle neoplasie del colon

New mathematical methods in medical diagnosis promise to bring great benefits to humanity, through achieving very high levels of specificity and sensitivity of the procedure. Moreover, as these methods can handle many variables, defined over different scales and can be used for synergic or non linear relationships a wealth of diagnostic analysis can be obtained from these methods, resulting in a high selectivity in categories of diagnosis. The aim of this paper is to show that a particular algorithm called T.R.A.C.E. (Total Recognition by Adaptive Classification Experiments) can be usefully applied in the histopathological analysis for diagnosis of colon cancer, with a very high sensitivity and an equally high specificity, which match the performance of the anatomo pathologist. In this way, through the formal structure required for this procedure, consisting of invariance, stability and pairwise exclusion of the diagnosis, a procedure could be defined for diagnosis free of subjective judgements and more in line with scientific principles. The outline of the paper is the following. After the introduction, in section 2 the algorithm is presented and distinguishing features compared to other pattern recognition methods are indicated. The data sample and the procedure to collect it and determine the appropriate patterns are examined in section 3, while in section 4 the results of the experimentations are stated, both for individual images and for the specimens. In section 5, certain generalisations and extension are described, while in section 6, conclusion follow.