海洛因成瘾为慢性复发性脑疾病，已成为重大社会问题，临床表现核心特征为复吸率高，高达90%，其潜在机制亟待深入研究，影像组学为复吸预测提供新思路。前期我们的神经影像学研究发现，毒品成瘾强制戒断患者存在脑结构和功能可塑性变化。在此基础上，首先对强戒的海洛因成瘾患者进行多时点纵向跟踪，考察非复吸组强戒疗效并结合多模态影像信息揭示其神经机制；然后融合不同模态磁共振影像数据，获取更加全面准确的影像学特征，构建基于全脑体素的海洛因成瘾预测模型；进一步借助连接度组学的研究方法，采用高空间分辨率的图谱，在大尺度复杂脑网络水平上，更加精确地挖掘出与海洛因依赖复吸行为密切相关的神经影像学靶点。最终尝试构建基于影像组学的最优复吸预测模型，用于监测和揭示海洛因成瘾人群疾病发展规律，提高海洛因成瘾复吸预测准确率，为强戒干预治疗及复吸预测提供临床支持。

Heroin addiction is a chronic recurrent brain disease, which has become a serious social problem. The core clinical manifestation of heroin addiction is high relapse rate, up to 90%. Its potential mechanism needs to be further investigated. Radiomics provide us a novel method to predict relapse. Our previous neuroimaging studies revealed the brain structure and functional plasticity changes in substance use disorders patients after compulsory withdrawal. Based on our previous findings, firstly, this project used long-term longitudinal design to explore the treatment outcomes of compulsory withdrawal and the possible neural mechanism in heroin-dependent patients by combining multimodality neuroimaging. Secondly, different modality magnetic resonance imaging (MRI) data are fused to derive more comprehensive and accurate imaging features, and build a whole brain voxel-based heroin addiction prediction model. Furtherly, we tried to reveal the more accurate neuroimaging targets at the level of large-scale complex brain networks by means of the connectome with a higher spatial resolution, which would be closely related to heroin-dependent relapse behavior. Finally, an optimal relapse prediction model based on radiomics is constructed to monitor and reveal the development of disease in heroin-dependent patients, improve the accuracy of relapse prediction, and provide clinical support for intervention therapy and relapse prediction.