| 陈岩岩,王艳昕,文曼,等.基于磁共振成像的影像组学在帕金森病抑郁诊断中的应用研究[J].安徽医药,2024,28(12):2407-2411. |
| 基于磁共振成像的影像组学在帕金森病抑郁诊断中的应用研究 |
| Application of magnetic resonance imaging-based radiomics in the diagnosis of depression in Parkinson's disease |
| |
| DOI：10.3969/j.issn.1009-6469.2024.12.015 |
| 中文关键词: 帕金森病 抑郁 影像组学 机器学习 预测模型 逻辑回归 |
| 英文关键词: Parkinson disease Depression Radiomics Machine learning Predictive models Logistic regression |
| 基金项目:安徽高校自然科学研究项目( KJ2020A0438)；新安医学教育部重点实验室开放基金项目( 2020xayx09) |
|
| 摘要点击次数: 500 |
| 全文下载次数: 266 |
| 中文摘要: |
| 目的根据磁共振成像( MRI)影像组学特征以及 4种机器学习算法来构建不同的机器学习帕金森病抑郁( dPD)预测诊断模型，以早期预测诊断 dPD。方法选取 2022年 10月至 2023年 11月安徽中医药大学第一附属医院明确诊断为帕金森病(PD)的 111例住院病人，根据量表及临床医生评估分为 PD组( n=64)和 dPD组( n=47)收集两组病人的临床资料、 MRI数据构建影像组学特征并进行特征筛选，根据所筛选的特征使用 4种不同的机器学习算法构建，dPD风险预测模型。对 dPD风险预测模型采用受试者操作特征曲线下面积( AUC)、准确度、灵敏度、特异度等指标评估模型效能。结果共筛选出 7个和 dPD密切相关的 MRI特征， 4种机器算法中逻辑回归( LR)算法所构建的模型效能最佳，训练集 AUC为 0.83、灵敏度为 0.86、准确度为 0.71；测试集 AUC为 0.77、灵敏度为 0.90、准确度为 0.79。结论基于 MRI特征和 LR算法所构建的 dPD预测诊断模型可以较为准确地诊断 dPD，为早期识别 dPD提供影像学指导。 |
| 英文摘要: |
| Objective To construct different machine learning risk prediction models for depression in Parkinson's Disease (dPD)based on the radiomics characteristics of magnetic resonance imaging (MRI) and four machine learning algorithms, so as to predict thediagnosis of dPD at an early stage.Methods A total of 111 inpatients with Parkinson's disease (PD) were selected from October 2022to November 2023 in the First Affiliated Hospital of Anhui University of Traditional Chinese Medicine, and were divided into PD group(n=64) and dPD group (n=47) according to the scale and clinician evaluation.The dPD risk prediction model was evaluated by using thearea under the ROC curve (AUC), accuracy, sensitivity, specificity and other indicators. Visualize the predictive model using a nomo di-agram.Results In this study, seven MRI features closely related to dPD were identified. Among the four tested machine learning algo-rithms, Logistic Regression (LR) performed the best. In the training set, LR achieved an AUC of 0.83, a sensitivity of 0.86, and an accu-racy of 0.71. In the validation set, the AUC was 0.77, the sensitivity was 0.90, and the accuracy was 0.79.Conclusion The dPD pre-diction and diagnosis model based on MRI features and LR algorithm can accurately diagnose dPD and provide imaging guidance forearly identification of dPD. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
