【摘要】：針對(duì)3D顱部醫(yī)學(xué)圖像配準(zhǔn)中存在的配準(zhǔn)精度不高、運(yùn)算復(fù)雜、配準(zhǔn)效率低等問(wèn)題,在創(chuàng)新性地圈定了感興趣配準(zhǔn)區(qū)域的基礎(chǔ)上,提出了一種基于Clifford代數(shù)的全新的幾何特征軸構(gòu)造方法。起初從參考模態(tài)與浮動(dòng)模態(tài)中依次提取特征點(diǎn),通過(guò)該特征點(diǎn)實(shí)現(xiàn)配準(zhǔn)感興趣區(qū)(ROI)圈定;其次利用感興趣區(qū)的點(diǎn)云數(shù)據(jù)集到其質(zhì)心的距離測(cè)度構(gòu)造幾何特征軸,并計(jì)算相應(yīng)的旋轉(zhuǎn)算子完成浮動(dòng)模態(tài)相對(duì)于參考模態(tài)的高效、高精度配準(zhǔn)。這樣的配準(zhǔn)方式有效地避免了多模態(tài)圖像成像時(shí)配準(zhǔn)區(qū)域非完全匹配導(dǎo)致的誤差,并減少待處理的數(shù)據(jù)量,同時(shí)消除了無(wú)效配準(zhǔn)區(qū)域產(chǎn)生的局部最優(yōu)點(diǎn)的影響,進(jìn)而降低了配準(zhǔn)的誤差。實(shí)驗(yàn)表明,感興趣區(qū)處理后的待配準(zhǔn)圖像,經(jīng)新算法仿真配準(zhǔn),能夠精確地定位組織器官的三維位置,執(zhí)行效率高且配準(zhǔn)誤差較小,是一種有效的3D顱部醫(yī)學(xué)圖像配準(zhǔn)方法。
[Abstract]:Aiming at the problems existing in 3D cranial medical image registration, such as low registration accuracy, complicated operation and low registration efficiency, the region of interest is delineated innovatively. A new method of constructing geometric feature axis based on Clifford algebra is proposed. At first, the feature points are extracted from the reference mode and floating mode in turn, and the region of interest (ROI) is registered by this feature point. Secondly, the geometric feature axis is constructed by using the distance measure from the point cloud data set of the region of interest to its centroid. The corresponding rotation operator is calculated to achieve the high efficiency and high precision registration of the floating mode relative to the reference mode. This registration method can effectively avoid the error caused by incomplete matching of registration region in multi-mode image imaging, reduce the amount of data to be processed, and eliminate the influence of local optimal points caused by invalid registration region. Furthermore, the error of registration is reduced. The experimental results show that the new algorithm can accurately locate the 3D position of tissue and organ, and the efficiency of the algorithm is high and the registration error is small. It is an effective method for the registration of 3D cranial medical images.
【作者單位】： 南通大學(xué)電氣工程學(xué)院;
【基金】：國(guó)家自然科學(xué)基金(61273024,61305031) 江蘇省自然科學(xué)基金(BY2016053-11) 江蘇省“333”高層次人才培養(yǎng)工程(BRA2015366) 江蘇省優(yōu)勢(shì)學(xué)科(PAPD)資助
【分類(lèi)號(hào)】：R741;TP391.41
，

本文編號(hào)：2174400