帕金森病(PD)是神经变性疾病，黑质内多巴胺能神经元退变、铁沉积增加、神经黑色素(NM)减少是重要的病理改变，其中黑质体-1（N1）受累最早、程度最重。N1铁含量增加导致MR图像上“燕尾征”消失现象对PD诊断有较高敏感性和特异性，但文献和申请人前期研究显示，N1的显示和MR序列选择、图像分辨率、信噪比及判读标准有密切关系。因此，基于“燕尾征”消失作为PD诊断标志物这一假说，申请人采用新的多模态MR序列STAGE开展志愿者预实验，在基于高分辨率幅度图和定量磁化率成像（QSM）的tSWI上能清晰显示NI，测得稳定的磁化率值。本项目拟在优化扫描的基础上，制定判读标准，创建形态学模板，设计多任务、模态注意力机制和深度学习网络对N1进行“分割-识别”，将N1结构、黑质内铁沉积及NM退变影像组学特征和深度学习网络高通量特征有机融合，建立基于影像标记物的PD诊断模型，为临床诊断提供快速易用的工具。

Parkinson's disease (PD) is a neurodegenerative disease. The main pathological changes of PD include dopaminergic neuron degeneration, iron deposition increase and neuromelanin (NM) reduction. Among them, substantia nigra-1 (N1) is the earliest and most severely affected substantia nigra structure. The phenomenon of disappearance of the "Swallowtail sign" on MR images due to the increase of N1 iron content has higher sensitivity and specificity for PD diagnosis. However, our and others previous studies have shown that the illustration of N1 is highly related to the selection of MRI sequences, image resolution, signal-to-noise ratio and interpretation criteria. Therefore, based on the hypothesis that the "Swallowtail sign" disappears as a PD diagnostic marker, our preliminary study on volunteers using a new multi-model MR sequence STAGE has showed that tSWI based on high-resolution phase images and quantitative susceptibility mapping can clearly show N1 and measure stable susceptibility values. Based on the optimal scan protocol, this project intends to formulate N1 structure interpretation criteria and create a N1 atlas. In addition, we are going to design multi-tasks and multi-models attention mechanism and deep learning network for the “segmentation and recognition” of N1 structure, and organically combine the radiomics features of N1 structure, iron deposition in substantia nigra and NM degeneration with the high-throughput characteristics of the deep learning network. Our ultimate goal is to establish diagnostic model of PD patients based on image markers and provide a quick and easy quantitative tool for clinical diagnosis of PD.