A PC-based medical image analysis system for brain CT hemorrhage area extraction

A PC-based medical image analysis system for extracting the brain CT hemorrhage area is developed. Two kinds of segmentation method are investigated: a fuzzy Hopfield neural net (FHNN) and a possibilistic Hopfield neural net (PHNN), for classifying the histogram of hemorrhage images, and the near-optimal threshold values can then be found. In addition, a function for professional doctors to extract the hematoma area is also included. The manual extraction areas are compared to those of the automatic extraction system and the average accuracy can be estimated. The FHNN is a neural net-based classification method for the fuzzy c-means algorithm. It has been proposed for automatic CT and MRI image segmentation in order to extract regions of interest. In addition, we have developed the PHNN as a new algorithm for classification. It has been developed by embedding the objective function of possibilistic c-means into the energy function of a modified Hopfield neural net. Thus, a near-optimal solution can be found by minimizing the energy function.