Research And Implementation On Process For Matching Huge Amount Of Medical Images

Matching medical image data is the key factor of the computer aided diagnosis. There are too much image data in database for doctor to check every time during the diagnosis. Matching image data by computer can help doctor to reduce the data to check, improve the efficiency of diagnosis. So for the past several decades, many image processing technologies have been developed and discussed, some methods like image matching base on gray value, based on Fourier Transfer, based on simulated annealing, based on wavelet and so on. However, most of the methods remain to be only of theoretical interests since the time complexity of the matching methods are too high to be realistic on handling huge amount of existing medical images.This paper presents a parallel processing model for matching huge amount of MR images. The parallel processing model is based on equal parallel model, and use database instead of network to transfer data, so the efficiency can be improved. Also the parallel model can be extended easily because it doesn't need control node. A feature vector of an MR image is defined by professionals specifically in the area of neuroscience, the feature vector including the feature that can show the feature of illness of the brain, such as area, gray value, shape, situation and so on. Then a matching algorithm is developed based on matching the feature vectors, the importance of the feature in feature vector is also defined by professionals specifically in the area of neuroscience, and its show in the algorithm as the weight of the feature. The experiments show that the overhead of synchronizing the parallel process is much less significant than the improvement of the overall efficiency.