基于ABUS图像的轻量型切口疝补片计算机辅助检测与评估算法

发布时间：2019-01-05 01:04

【摘要】：由于人工检阅自动化三维乳腺超声(ABUS)图像极其耗时,而且极易出现对微弱异常区域的漏诊.为了提高ABUS图像检阅效率并减少漏诊,提出一套基于ABUS图像的轻量型切口疝补片计算机辅助检测与评估算法.算法首先使用纹理特征萃取算法自动量化提取三维感兴趣容积中的待分类区域的相关纹理特征参数,以便用于对补片和筋膜的区分;然后,针对二维纹理参数对切口疝补片术后卷曲、收缩等空间变换较为敏感的问题,引入三维纹理参数和三维位置参数来提高轻量型补片分类识别算法的鲁棒性;最后,使用类间距算法和顺序前进搜索法来进行特征选择,并使用支持向量机进行分类识别.算法可有效降低人工阅片的工作强度,辅助医生识别ABUS扫描区域内有无轻量型补片,并对补片相关诊断项目做出辅助评估.
[Abstract]:Due to the time consuming of automatic 3D ultrasound (ABUS) images, it is easy to misdiagnose the weak abnormal areas. In order to improve the efficiency of ABUS image inspection and reduce missed diagnosis, a set of computer-aided detection and evaluation algorithm for light incision herniation patch based on ABUS image was proposed. Firstly, the texture feature extraction algorithm is used to automatically extract the relevant texture feature parameters of the region to be classified in the three-dimensional volume of interest in order to distinguish the patch from the fascia. Then, aiming at the problem that two-dimensional texture parameters are more sensitive to space transformation such as crimp and contraction after incision herniation, 3D texture parameters and 3D position parameters are introduced to improve the robustness of the classification and recognition algorithm of lightweight patch. Finally, feature selection is carried out using class spacing algorithm and sequential forward search method, and support vector machine is used for classification and recognition. The algorithm can effectively reduce the working intensity of manual film reading, assist doctors to identify whether there are lightweight patches in the ABUS scan area, and make an auxiliary evaluation of the diagnostic items related to the patch.
【作者单位】：昆明医科大学第一附属医院科教部;天津大学微电子学院;云南大学信息学院;复旦大学附属华东医院超声科;
【基金】：国家自然科学基金(61661050) 云南省教育厅科学研究基金重点项目(2015Z013) 上海申康医院发展中心临床辅助科室(超声医学)能力建设项目(SHDC22015013)
【分类号】：R656.2;TP391.41

【相似文献】