Study On Medical Image Processing And Recognition Method

Study on medical image processing and recognition of research has obviously importanttheoretical significance and practical value. The methods of segmentation, fusion andrecognition for thyroid CT image and human brain image were studied in order to achieve thecomputer-aided diagnosis and to provide reference and help for the doctor’s clinical diagnosisand treatment.Firstly, the histogram modification, denoising and sharpening processing methods areused to preprocess the thyroid CT images and human brain images to make these images easyfor machine description and analysis.Secondly, an improved image segmentation algorithm is proposed based on the principleof various image segmentation algorithms, according to the characteristic of the thyroid image.For the first step, the image edge is extracted by using Sobel operator with8directions’templates. For the second step, the improved method proposed by improving the originaliterative threshold segmenting method is used for the threshold segmenting to getsegmentation images.Then, an image fusion algorithm of the threshold image segmentation method combinedwith the image edge detection method is provided according to the requirement of the medicalimage analysis and processing. The new algorithm is given to obtain the fused image which isof abundant details information and smooth edge and easy for computer to describe. Thefused image helps doctor to diagnosis.Finally, a new kind of feature is defined based on the area ratio and a classificationmethod is presented. After the classified features are extracted, the K-nearest neighborclassification algorithm is adopted to classify thyroid CT images that belonged to the normalor the abnormal categories.The experimental results verify that the boundaries of the segmented images and thefused images are continuous, smooth. Compared with the original boundary detection and iterative threshold segmentation algorithm, the improved algorithm is more effective insegmentation. Moreover, the experimental results also show the new feature is effect and therecognitions of the normal and the abnormal thyroids are correct. Then, the new way ofcomputer-aided diagnosis of thyroid is achieved.