CASIA OpenIR  > 中国科学院分子影像重点实验室
乳腺癌近红外激发荧光成像方法研究和临床应用探索
张崇
Subtype博士
Thesis Advisor田捷
2020-05-24
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword近红外激发荧光成像 手术导航 图像处理 乳腺癌
Abstract

乳腺癌是世界三大高发癌种之一,尤其在女性患者中,患癌发病率甚高。传统的乳腺癌治疗手段是手术切除,但根治性切除术式,切除范围大,创口大愈合难,所以保乳手术被越来越多的乳腺癌患者接受。但保乳术式存在术中病灶识别主观性强的问题,由此可能造成的阳性切缘会阻碍手术进程,甚至影响患者的术后生存期和生活质量。此外,病理检查时,不可触的乳腺癌会给医生带来识别癌灶的挑战。因此,急需一种客观的乳腺癌成像方法来辅助医生做出临床决策。近红外荧光手术导航技术是一种新兴的术中成像手段,其通过造影剂汇聚在病变区域和重要组织处,激发光照射,发射出的信号被设备探测到,从而实现实时动态的组织荧光成像。但近红外荧光成像因组织散射和光信号缺失、设备参数限制等问题,会造成成像效果模糊。基于以上问题,本文立足于近红外荧光成像分辨率增强算法的研究,和乳腺癌近红外激发荧光成像设备在临床应用的探究,开展了相应的工作。论文的主要工作和创新点归纳如下:

1.在成像方法方面,提出了一种基于GAN的近红外荧光图像分辨率增强算法,使低质模糊的荧光图像,经处理后边缘锐化,实现图像的分辨率增强。针对荧光成像的组织散射和设备参数限制等导致的荧光成像模糊的问题,研究近红外荧光的图像后处理方法,提出基于GAN网络的荧光图像分辨率增强算法。算法创新点在于网络结构的对称化设计,训练数据集和训练过程的设计,以及提出全梯度损失函数用于网络训练,进一步优化网络结构。将训练出的网络应用于分辨率板测试,小鼠尾血管成像和临床乳腺癌淋巴管示踪成像上,实验结果表明了该算法可以实现原始荧光图像的分辨率增强。

2.在成像设备方面,搭建了一套适用于美蓝谱段的、可以成像乳腺癌的近红外激发荧光成像设备,并对设备性能进行测试。针对目前国内缺少可以在700nm近红外谱段、实时成像乳腺癌的近红外荧光成像设备,这一关键科学问题,设计了700nm谱段的荧光成像设备,结合美蓝造影剂,以实现乳腺癌成像。基于实验室前期手术导航设备的开发经验和设备基础,设计设备的信号采集和成像单元,使其适用于美蓝的光学谱段成像。完成设备搭建后,对美蓝浓度和成像分辨率分别进行测试。实验结果表明该设备在四个美蓝试剂的浓度梯度中,较好的成像浓度为0.01mg/ml,成像分辨率接近期望水平,满足设备对美蓝造影剂的成像标准。

3.在成像应用方面,将搭建的乳腺癌近红外激发荧光成像设备应用于临床,探究该设备对人体离体乳腺癌组织的成像效果,并病理对照,统计和分析。本次试验不影响手术常规流程,患者术前3小时静脉注射1mg/kg美蓝试剂,行乳腺切除术后,离体标本送往病理室,进行近红外激发荧光成像。在入组的30例患者中,20例非术前化疗患者,实现乳腺癌荧光成像16例;10例行术前化疗患者中,乳腺癌成像3例。除30例肿瘤组织外,有额外5处可疑组织有荧光显影,病理检查,证明是良性组织伴有出血点。因此,本文用美蓝荧光成像方法检测乳腺癌的灵敏度为0.63,阳性预测值为0.79。试验结果表明自主研发的设备对乳腺癌成像的可行性,但术前化疗会降低患者乳腺癌组织的显影比例。

本研究围绕乳腺癌近红外激发荧光成像这一课题,提出了一种用于荧光成像的分辨率增强算法,搭建了一套基于MB、适用于乳腺癌成像的激发荧光成像设备,并开展了乳腺癌激发荧光成像设备在临床应用上的可行性研究。相关研究结果证明了算法和设备的可行性,以及美蓝荧光成像乳腺癌的可行性。

Other Abstract

Breast cancer is one of the three most common cancers in the world and is especially a common malignancy in women. Traditional treatment for breast cancer is surgery. However, the radical resection method leads to a large wound, which is often difficult to heal. Therefore, breast-conserving surgery (BCS) is being accepted by more and more breast cancer patients. During the BCS, surgeons removed the cancerous lesion with a certain degree of subjectivity, which may lead to positive resection margin. It would hinder the surgical process, and even affect the patient survival and quality of life. In addition, non-palpable breast cancer presents a challenge to identify the lesion for pathologists during pathological examination. Therefore, an objective breast cancer imaging method is urgently needed to assist doctors in making clinical decisions. Near-infrared fluorescence surgical navigation technology is an emerging intraoperative imaging method. Contrast agent accumulates in specific tissue. When illuminated by an excitation light, the contrast agent with various concentrations in tissues emits a fluorescent signal, which is received by a fluorescence imaging equipment for real-time imaging. However, near-infrared fluorescence imaging can cause blur due to problems such as light signal scattering and loss, as well as equipment parameter limitations. Based on the above, this paper has carried out research on the resolution enhancement algorithm of fluorescence imaging and the clinical application of breast cancer near-infrared fluorescence imaging equipment. The main work and innovations of this article can be summarized as follows:

1. In terms of imaging method, a fluorescence image resolution enhancement algorithm based on generative adversarial network (GAN) is proposed. Low-quality and blurred fluorescence images are sharpened after processing to achieve resolution enhancement. Aiming at the problems of blurred images caused by light signal scattering and equipment parameter limitation of fluorescence imaging, the post-processing method of fluorescence images is studied, and a fluorescence image resolution enhancement algorithm based on GAN is proposed. The innovation of the algorithm lies in the symmetrical design of network structure, the design of the training data set and the training procedure, as well as the total gradient loss, which is proposed for networking training to further optimize the network structure. After training, the network was applied to the resolution plate test, vessels imaging experiment in mouse tail and lymphatic imaging of breast cancer in clinical trial. Experimental results show that the algorithm can achieve the resolution enhancement of original fluorescence images.

2. In terms of imaging equipment, a set of near-infrared fluorescence imaging equipment, which is suitable for MB spectrum and can realize breast cancer imaging, is developed and the performance of the equipment is tested. In order to solve the problem of lacking domestic real-time fluorescence imaging equipment for breast cancer visualization in the near-infrared band of 700nm, a MB-specific equipment with 700nm spectrum characteristic was designed, which could realize breast cancer imaging combined with MB. Based on the previous experiences for the development of surgical navigation equipment in the laboratory, the signal acquisition and imaging unit of the equipment is designed to be suitable for the imaging effect of MB optical spectrum. After the equipment completed, the MB concentration test and the imaging resolution test were performed separately. The test results show that concentration of 0.01mg/ml has a better imaging effect among four MB gradient concentrations, and the imaging resolution is close to expected level, which meets the MB imaging standards of the equipment.

3. In terms of imaging applications, the home-made near-infrared fluorescence imaging equipment was applied to clinical application for breast cancer visualization in resected human tissues and the imaging effect were explored through pathological statistics and analysis. The experiment did not affect the routine procedure of surgery. Patients were injected with 1mg/kg MB intravenously 3 hours before the surgery. After resection, the specimen was sent to the pathology room for fluorescence imaging. Among the 30 patients enrolled, 16 of 20 non-preoperative chemotherapy patients achieved fluorescence signal detection in breast cancer area of specimen. In contrast, 3 of 10 patients with preoperative chemotherapy were detectable. Besides 30 tumor samples, 5 more suspicious samples with fluorescence signal were confirmed to be benign hemorrhagic tissues. Therefore, a sensitivity of 0.63 and a positive predictive value of 0.79 were achieved by the methylene blue fluorescence imaging strategy. The results show the feasibility of the home-made MB-based near-infrared fluorescence imaging equipment for breast cancer visualization, but preoperative chemotherapy may reduce the detection rate.

This study focuses on the near-infrared fluorescence imaging technology for breast cancer visualization. A resolution enhancement algorithm for fluorescence images was proposed; a MB-based fluorescence imaging equipment for breast cancer visualization was developed, and a clinical feasibility study for breast cancer visualization by using this home-made equipment was carried out. The relevant studies have proven the feasibility of the algorithm and the equipment, as well as the MB-based fluorescence imaging technology for breast cancer visualization.

Pages113
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/39019
Collection中国科学院分子影像重点实验室
Recommended Citation
GB/T 7714
张崇. 乳腺癌近红外激发荧光成像方法研究和临床应用探索[D]. 中国科学院自动化研究所. 中国科学院大学,2020.
Files in This Item:
File Name/Size DocType Version Access License
乳腺癌近红外激发荧光成像方法研究和临床应(5418KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张崇]'s Articles
Baidu academic
Similar articles in Baidu academic
[张崇]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张崇]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.