Intelligent Data Mining Algorithms Research And Their Application In The Medical Image Area

This research work is a part of the Research of Data Mining on Medical Image supported by the National Natural Science Foundation of China. Based on the intelligent data mining algorithm, some key techniques such as NN(Neural Network), Rough Set and so on have been studied, and a computer-aided medical diagnosing system on breast cancer has been developed. The main research work include:1. The designing,comparison and implementation of the BP(Back-Propogation) and RBF(Radial Basis Function) Neural Network. Experimental results prove that the RBF NN performs better than BP NN in terms of classification accuracy.2. Lucubrating in RS(Rough Set) TheoryIn the system, an integrated algorithm used to analysis slice up data (from UCI database) is proposed and implemented, which is based on the attributes reduction and rules extraction, obtaining the core and effective rules, which are more comprehend than original data, and can help the doctor to understand patient's state of the illness quickly.The reduced attributes are obtained using RS without former knowledge, and they are totally based on the data in database with no relation to the value of data, so it is highly suitable for discrete data. Then the continuous data must be discretized before using RS theory on them. Several discretization algorithms have been discussed and implemented. Discretization algorithms affect the reduction attributes to some extent.3. Image enhancement and shape Features ExtractionAccording to the characteristics of the medical image, the image enhancement algorithm based on the RS theory is improved and used in the medical area for the first time in this system. Comparison results illustrate that the enhancement effect based on RS Theory is obvious and better than the histogram equalization.A region growing method to extract accurate boundary of the breast tumor region was investigated and implemented. Shape features are discussed and three shape features (Compactness, Fourier Descriptor, and Improved Moment Invarient)have been extracted. Experimental results show that the shape features representthe shape of breast tumor perfectly and is very effective in distinguishing thebenign from the malignant tumor.4. The computer-aided medical diagnosing system is designed and implemented, which has friendly interface, and can be easily expanded. The system makes it possible that automatic diagnosis of breast cancer between normal and abnormal, benign and malignant. The accuracy is relatively high based on a standard MIAS Database.