CONSTRUCTION AND APPLICATION OF HIERARCHICAL DECISION TREE FOR CLASSIFICATION OF ULTRASONOGRAPHIC PROSTATE IMAGES

A non-parametric algorithm is described for the construction of a bina ry decision tree classifier. This tree is used to correlate textural f eatures, computed from ultrasonographic prostate images, with the hist opathology of the imaged tissue. The algorithm consists of two parts; growing and pruning. In the growing phase an optimal tree is grown, ba sed on the concept of mutual information. After growing, the tree is p runed by an alternating interaction of two data sets. Moreover, the st ructure and performance of the constructed tree are compared to the re sults using a slightly modified corresponding growing and pruning algo rithm. The modified algorithm provides better retrospective and prospe ctive classification results than the original algorithm. The use of t he tree for automated cancer detection in ultrasonographic prostate im ages results in retrospective and prospective accuracy of 77 . 9% and 72 . 3%, respectively. Using this tissue characterisation, a supportin g tool is provided for the interpretation of transrectal ultrasonograp hic images.