Biomedical Data Classification Research Articles

Computerized consensus diagnosis: a classification strategy for the robust analysis of MR spectra. I. Application to 1H spectra of thyroid neoplasms.

We introduce and apply a new classification strategy we call computerized consensus diagnosis (CCD). Its purpose is to provide robust, reliable classification of biomedical data. The strategy involves the cross-validated training of several classifiers of diverse conceptual and methodological origin on the same data, and appropriately combining their outcomes. The strategy is tested on proton magnetic resonance spectra of human thyroid biopsies, which are successfully allocated to normal or carcinoma classes. We used Linear Discriminant Analysis, a Neural Net-based method, and Genetic Programming as independent classifiers on two spectral regions, and chose the median of the six classification outcomes as the consensus. This procedure yielded 100% specificity and 100% sensitivity on the training sets, and 100% specificity and 98% sensitivity on samples of known malignancy in the test sets. We discuss the necessary steps any classification approach must take to guarantee reliability, and stress the importance of fuzziness and undecidability in robust classification.

SEQUAL: an interactive computer program for sequential classification of biomedical data

A computer program for sequential bayesian classification of patterns defined by integer and real-valued data is described. Classified patterns from a training sample are used to estimate the non-parametric (kernel) probability density functions and the a-priori class probabilities necessary to implement the bayesian classification. For each pattern and at each step in the sequential program, the ‘best’ feature to be measured at the next step is computed on the basis of the estimated misallocation error rate. The user can actually use the proposed feature or any other one; once the chosen feature has been measured, its value is used to allocate the pattern into the class with the highest conditional a-posteriori probability, according to the Bayes formula. The main feature of the program consists in the computation of the ‘probability of reversal’ at each step of the sequential procedure. The probability of reversal represents the probability that at the next step the pattern will be classified into a class different from the present one. The probability of reversal can be used as a stopping criterion, which is more efficient than other commonly used stopping rules, such as the a-posteriori Bayes probability or the estimated misallocation error rate. The program, available in FORTRAN 77 for a VAX/VMS machine, has been tested both on simulated and real data collected from patients suffering from various forms of hepatic disease.