基于图像和分子信息的胃黏膜上皮内瘤变无创量化分级与检测研究

The 5 year survival rate of early gastric cancer is much higher than that of advanced gastric cancer. Intraepithelial neoplasia of gastric mucosais is recognized as the most direct precancerous lesion of gastric cancer. Its early detection and grading can improve the detection rate of early gastric cancer. Endoscopy is a routine method for detecting this lesion. However, it is difficult for physicians to classify the severity of intraepithelial neoplasia by endoscopy. It is easy to lead to the missed diagnosis of this lesion in subsequent biopsies. In addition, the accuracy of endoscopic diagnosis is influenced by the physician's subjectivity, experience and visual cognitive ability. The extent of the patient's willingness to be diagnosed with endoscopy is low. On the basis of previous work, this research aims to solve the quantitative grading of intraepithelial neoplasia with endoscopic diagnosis and constructs a accurate in vitro intelligent detection model for intraepithelial neoplasia, using the technologies of medical image processing, microarray, bioinformatics and deep learning, etc. The research contents include: endoscopic image-based intraepithelial neoplasia detection and 3D reconstruction of pathological mucosa; quantitative grading of intraepithelial neoplasia based on 3D image feature extraction; screening and functional study of exocrine derived microRNA markers associated with high-level intraepithelial neoplasia; construction of high level intraepithelial neoplasia detection model based on combinatorial molecular markers. The study results will provide new ideas and new methods to increase accuracy of detecting this disease and are helpful to understand its progression mechanism. Therefore, it has a very important scientific significance.

早期胃癌的5年生存率远远高于晚期胃癌。胃黏膜上皮内瘤变是被公认的胃癌最为直接的癌前病变，尽早发现和明确分级可以提高早期胃癌的检出率。内镜是诊断该病变的常规手段。然而，医师难以用内镜对上皮内瘤变严重程度分级，容易导致后续组织活检漏诊。内镜诊断准确率受到医师主观性、经验和视觉认知能力的影响，且患者接受程度低。本研究将在前期工作基础上，拟采用医学图像处理、微阵列、生物信息学和深度学习等技术，解决上皮内瘤变在内镜诊断环节的量化分级问题，建立一种准确的上皮内瘤变体外智能检测模型。研究内容包括：基于内镜图像的上皮内瘤变检测与病变黏膜3D重构；基于3D图像特征提取的上皮内瘤变量化分级；高级别上皮内瘤变相关的外泌体源性microRNA标志物筛选及功能研究；基于组合分子标志物的高级别上皮内瘤变检测模型构建。研究成果将为提高该病变检测准确度提供新思路和新方法，将有助于认识其进展机制，具有十分重要的科学意义。

早期胃癌的5年生存率远远高于晚期胃癌，但其临床诊断准确率受到医师主观因素的影响较大，容易发生漏诊和误诊。因此，本项目的研究目标是要采用医学图像处理、微阵列、生物信息学和深度学习等多学科技术，解决从癌前病变到早期胃癌在内镜诊断环节的量化分级问题，建立一种准确的胃部高级别上皮内瘤变和早期胃癌体外智能检测模型。. 围绕着本项目的研究目标，研究团队将全部研究任务分解为4项研究内容，已经完成（i）提出了3种基于内镜图像的上消化道肿瘤检测与分割方法；（ii）建立了2种基于内镜图像特征提取的上消化道病变量化分级方法;（iii）从非编码RNA协同互作和竞争网络与模块的不同角度，开展了恶性肿瘤相关的非编码RNA标志物筛选及协同调控模式研究;（iv）建立了2种基于分子标志物的胃部肿瘤检测与预后风险预测模型。此外，本研究团队还将研究对象从早期胃癌/癌前病变拓展到早期食管癌和肠癌，也将研究拓展到了肿瘤心脏病学这一新兴领域，为后期开展消化道肿瘤治疗过程中心脏毒素研究奠定了坚实基础。. 本项目取得的研究成果为提高消化道早期癌症诊断准确率提供了新思路和新方法，对消化道早癌发生与发展机制提供了新的理解与认识，很好地达到了研究目标，具有十分重要的科学意义和实际临床应用前景。. 本项目支持14篇SCI论文发表和8项中国发明专利被授权（其中1项专利转化），培养4名博士后出站、7名博士生获得博士学位和6名硕士生获得硕士学位，支持2项硕士生参与的本项目研究成果分别获得第六、七届全国大学生生物医学工程创新设计竞赛一等奖和三等奖，本项目主研甘涛从副主任医师成长为主任医师。