Neural network and conventional classifiers for fluorescence-guided laser angioplasty

The ability of the back-propagation and K-nearest-neighbors techniques to classify arterial fluorescence spectra is investigated. Both methods are competitive with other classification schemes. The best validation set accuracy on the aortic data was obtained by the 1-nearest-neighbor method (98% correct overall on the test exemplars). The 22-8-1 and 22-8-4-1 networks performed second best, misclassifying only one more exemplar, at 96%. All performances on the coronary data were much poorer. The relative performance of variations on both techniques is used to make inferences about the geometry of the classification task.