基于超声-磁共振融合的影像组学技术对IDH1/2突变型胶质瘤术中分子边界可视化的初步研究
批准号:
82072020
项目类别:
面上项目
资助金额:
55.0 万元
负责人:
史之峰
依托单位:
学科分类:
治疗计划、导航与机器人辅助
结题年份:
2024
批准年份:
2020
项目状态:
已结题
项目参与者:
史之峰
中文摘要
胶质瘤是最常见的恶性脑肿瘤,最大范围安全切除是延长患者生存时间的首选疗法。由于胶质瘤呈浸润性生长,常规术中导航(MRI)仅能指导影像学全切而非分子全切。术中快速诊断IDH1/2突变能够帮助外科医生判定胶质瘤分子边界,但现有诊断技术尚不能实现分子边界的可视化,无法用于术中导航。课题组前期采用基于深度学习的影像组学技术对胶质瘤IDH1/2突变进行无损预测,准确率达到93%,进一步对MRI数据进行模块化处理,实现了IDH1/2突变的可视化和定量化。本课题将在此基础上深入研究术中超声与术前磁共振的融合导航及基于多模态影像组学的分子可视化技术,通过解决跨模态配准和转化重建的技术难点,纠正术中 “脑漂移”造成的MRI导航误差,验证基于IDH1/2分子边界可视化的准确性,最终获得精准、实时的胶质瘤分子边界可视化导航,帮助外科医生实现胶质瘤分子水平全切除,最大限度延长患者总体生存时间。
英文摘要
Glioma is the most common malignant brain tumor in central nervous system. Maximal safe resection is the preferent treatment to prolong patient survival time. Due to its infiltrative growth pattern, regular image-based neuro-navigation (MRI) is out of feasibility to guide molecular total resection. Rapid intra-operative IDH1/2 mutational status detection can be applied to delineate glioma molecular boundary. However, there is no relevant technology or method can be successfully established to visualize the entire tumor molecular boundary. In our previous study, we used a deep-learning based radiomics method to non-invasively predict IDH1/2 mutational status with accuracy of 93% by conventional MR images, furthermore, realizing IDH1/2 mutational visualization and semi-quantification of entire tumor. Hereafter, we plan to focus on in-depth research of developing modified radiomics-based molecular visualization method to delineate glioma molecular boundary by using combined ultrasound-MRI navigation. A novel modal conversion strategy is designated to resolve the challenge of trans-modality registration in neuro-navigation. Thus, combined ultrasound-MRI radiomics method can facilitate the accuracy of intra-operative molecular boundary delineation. Meanwhile, the application of ultrasound can correct the navigation error of “Brain Shift”. This novel method has great potential to help neurosurgeon visualize glioma molecular boundary in a real-time and precise manner so as to achieve molecular total resection. Patients will be benefit with prolonged survival time and more opportunities to treated with comprehensive medical healthcare.
胶质瘤是最常见的恶性脑肿瘤,因其浸润性生长常规MRI导航只能实现影像学全切而难以达到分子层面的精准切除。本项目针对胶质瘤术中快速分子诊断与预后预测需求,基于术中超声射频信号、术前磁共振影像和病理检测等多模态数据,构建了高质量的脑肿瘤术中超声-NGS数据库与术前磁共振-HE病理数据库,为胶质瘤多分子标志物实时诊断和治疗决策提供了关键数据支撑;在此基础上提出了融合Bi-LSTM、双域图卷积及内外图注意力机制的STIM模型,经过数据预处理(信号裁剪、归一化、分块)后实现了超声射频信号的时间与空间维度特征挖掘,并利用标签平滑正则化增强模型的泛化能力,从而在3秒内完成IDH1、TERTp、1p/19q等分子标志物的检测并通过热图可视化肿瘤异质性,帮助外科医生根据分子边界进行精准切除;同时,针对高级别复发风险的低级别胶质瘤患者,研究团队融合宏观MRI影像与微观病理切片信息,提出MFR多模态放射组学模型,通过提取和筛选放射组学特征及细胞核形态特征,并采用注意力机制进行加权融合,实现对替莫唑胺化疗敏感性的高精度预测,支持临床个性化治疗决策。上述研究成果为胶质瘤分子水平全切除及个性化治疗策略的优化提供了重要的技术手段和临床价值,也为后续深入探究胶质瘤浸润机制和提升患者整体生存率奠定了良好基础。
ATRX突变调控肿瘤相关免疫微环境诱导TMZ化疗后IDH突变型低级别胶质瘤恶性转化的机制研究
-
批准号:82373018
-
项目类别:面上项目
-
资助金额:49万元
-
批准年份:2023
-
负责人:史之峰
-
依托单位:
基于超声-磁共振融合的影像组学技术对IDH1/2突变型胶质瘤术中分子边界可视化的初步研究
-
批准号:--
-
项目类别:--
-
资助金额:55万元
-
批准年份:2020
-
负责人:史之峰
-
依托单位:
PLCG1基因表达调控IDH1/2野生型较低级别胶质瘤发生发展及其靶向治疗的基础研究
-
批准号:81702471
-
项目类别:青年科学基金项目
-
资助金额:20.0万元
-
批准年份:2017
-
负责人:史之峰
-
依托单位:
国内基金
海外基金