摘要
目的:设计基于磁共振成像(MRI)的膀胱肿瘤自动检测系统,为克服光学膀胱镜检测膀胱肿瘤有创及存在视觉盲区等缺点,寻找无创和便捷的早期检测路径。方法:根据建立的膀胱肿瘤自动检测方法,利用10套膀胱癌患者T_2加权影像序列数据进行初步验证,读取T_2加权MRI数据,使用自适应形状先验约束的水平集方法(ASPCLS)进行膀胱内外壁分割,利用拉普拉斯偏微分方程计算三维膀胱壁厚度,并以此为基础通过基于空域信息的模糊C均值(s FCM)算法提取肿瘤疑似区,最终完成提取膀胱肿瘤疑似区域的系统功能。结果:验证结果表明,所建立的系统能够较为准确的提取出膀胱肿瘤疑似区域,且与医生勾画的真实膀胱肿瘤区域的平均Dice重合系数(DSC)为86.7%,效果良好。结论:基于MRI的膀胱肿瘤自动检测系统,能够初步较准确地实现膀胱肿瘤的检测和疑似肿瘤区的提取,为膀胱肿瘤的早期检测及术前诊断提供无创新手段。
Objective To design automatic detecting system for bladder tumor based on magnetic resonance imaging （MRI） so as to overcome the disadvantages of invasive wound and blind area of optical cystoscope and explore early thinking for noninvasive and convenient detection. Methods： According to the established automatic detection method of bladder tumor, the T2 weighted image sequential data of 10 patients with bladder tumor were implemented primary verification. The T2-weighted MR images were acquired, and the adaptive shape prior constrained level sets （ASPCLS） algorithm was applied to cut inner and outer surfaces of the bladder wall. The three-dimensional bladder wall thickness（BWT） was calculated by using the Laplace partial differential equation. Based on above results, through used fuzzy C-Means with spatial information（sFCM） to extract suspected tumor regions. At finally, the systematic function of extracting suspected region of bladder tumor was realized. Results： The results indicated that the system could relatively accurate extract suspected region of bladder tumor, and the average Dice＇ similarity coefficient （DSC） of the suspected region drawn by system and the true region drawn by doctor was 86.7%. And it was better results. Conclusion： The automatic detection system can primarily realize the relatively accurate detection for bladder tumor and the extraction of suspected region. And it can provide newly noninvasive method for the early detection and preoperative diagnosis of bladder tumor.
出处
《中国医学装备》
2018年第1期2-5,共4页
China Medical Equipment
基金
国家自然科学基金重点项目(81230035)"基于影像定量分析和特征可视化的虚拟内窥镜关键技术研究"
陕西省社会发展项目(2016SF302)"基于多模态MRI膀胱癌定量影像特征的膀胱癌基因特性预测系统的开发"
关键词
膀胱肿瘤
磁共振成像
自动检测系统
肿瘤疑似区提取
Bladder tumor
Magnetic resonance imaging
Automatic detecting system
Extraction of suspected tumor region
