中文摘要：

目的对婴幼儿脑的64层CT低分辨率扫描图像进行分形计算及优化分析。方法将Toshiba Aquilion64层CT使用10mAs扫描婴幼儿脑的CT成像原始数据输入Matlab7．1图像工具箱，进行多重分形谱分析和实验性图像降噪，并与常规低剂量扫描50mAs组图像质量进行对比。结果10mAs原始图像的严重噪声导致医学诊断价值丧失，使用多重分形模型降噪后的图像具有良好的图像细节保持特性及良好的细节保持特性，虽然图像质量评分仍远不如常规剂量图像，但与原始噪声图像比较，其差异有统计学意义（F=38．85，P〈0．01），表明经分形模型降噪优化后的图像可以基本满足临床诊断。结论多重分形谱降噪可用于低剂量低分辨率CT图像的优化，能提高病变检出敏感性的对比度／噪声比，进一步研究应用有望大幅度减少婴幼儿CT扫描的辐射剂量。

英文摘要：

Objective To analyze scanned image optimization based on the muhifractal soectrum and image fractal algorithm of 64-slice spiral CT in brain of infant. Methods The image data of Toshiba Aquilion 64-slice CT scanning using 10 mAs were imported to image processing toolboxs of Matlab 7.1. The evaluation of muhifractal spectrum and image denosing were performed, and compared with image quality of conventional lowdose CT using 50 mAs. Results The low-contrast scanned image used 10 mAs is the valueless medical image because of serious noise. Image denoise based on the fractal model bad superior characteristic of image detail preserving and better contrast-to-noise ratio（CNR）. There existed a group difference in the score of image quality between the rude imaging noise and optimized image based on the muhifractal spectrum algorithm, though the score was still significantly lower than the normal dosage scanned image（ F = 38.85, P 〈 0.01）. The group difference was also manifested the image quality of infants can achieve basically the request of clinical diagnosis by suitable model denoising algorithm. Conclusions Image denoising based on the muhifractal spectrum model can be used on the low-dose and low-contrast CT image optimization. It improved the CNR of the pathological region. The radiation dose of CT scanning in infants would be declined significantly by its further application in the future.