Research And Application Of Medical Image Pattern Recognition Technology

Pattern recognition technology is based on the recognition mechanism of human brain, it uses computer to describe and classify various of things and their chance process.Medical image pattern recognition is an important branch of pattern recognition technology. With a large number of new equipment used in clinical, there are more and more kinds of medical images. This undoubtedly brings heavy work to doctors if it relies on subjective judgment. Thus, it is imperative to study the automatic recognition of medical image. Medical images have many differences from other images, such as more features, lower resolution and bigger correlation, and it needs to ensure the reliability of the diagnosis. Therefore we need to improve and perfect the existing algorithms according to the features of medical images to make them more suitable for medical image processing.In this dissertation, it has studied the pattern recognition in medical image processing, optimized the traditional recognition methods, and realized the recognition of the Meibomian gland morphology and the breast neoplasms. The specific research contents are as follows:(1) It has studied Several traditional image pattern recognition algorithms,analyzed their performance at certain conditions, and selected a more suitable method for medical recognition.(2) It has studied a recognition method based on improved FCM and rough set theory. This method combines the advantages of FCM and rough sets.It can reduce the redundant attributes and extract the most representative rules under the premise of ensuring the ability of classification. At the same time, it uses FCM to make the attribute fuzzy rather than discrete to effectively avoid information losses and make the decision rules more accurate. Besides, it proposes a new method based on distance to select the initial cluster centers, in witch the outliers are treated separately, so as to overcome the defect of the traditional FCM algorithm. The improved algorithm is used to recognize the meibomian gland morphology, and it reduces the computation and ensure the accuracy, so that the performance of the system is greatly improved.(3) It had studied the support vector machine and optimized its parameters by adaptive genetic algorithm. SVM has many advantages in pattern recognition,but the choice of parameters is difficult. This paper uses adaptive genetic algorithm to optimize the support vector machine. It avoids the traversal of all parameters, reduces the amount of computation greatly and make the search faster as well as the accuracy higher. The method is used to recognize breast cancer, by extracting the shape features of the tumor from the breast image which is pre processed, and inputting the feature vector to the SVM trained well.Is recognition rate can be 99.67%.(4) Based on the application of the above algorithm,we design a recognition system for medical image on the GUI platform of MATLAB, It includes the interface design and the program design for meibomian gland morphology recognition subsystem and breast tumor recognition subsystem, realizes the interaction between human and computer.