Li Jingwen, Liu Xiaoming, Wang Xinyi, Liu Hanshu, Lin Zhicheng, Xiong Nian
Department of Neurology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430022, China.
Department of Radiology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430022, China.
Brain Sci. 2022 Jun 29;12(7):851. doi: 10.3390/brainsci12070851.
Diagnosis of Parkinson's Disease (PD) based on clinical symptoms and scale scores is mostly objective, and the accuracy of neuroimaging for PD diagnosis remains controversial. This study aims to introduce a radiomic tool to improve the sensitivity and specificity of diagnosis based on Diffusion Tensor Imaging (DTI) metrics.
In this machine learning-based retrospective study, we collected basic clinical information and DTI images from 54 healthy controls (HCs) and 56 PD patients. Among them, 60 subjects (30 PD patients and 30 HCs) were assigned to the training group, whereas the test cohort was 26 PD patients and 24 HCs. After the feature extraction and selection using newly developed image processing software Ray-plus, LASSO regression was used to finalize radiomic features.
A total of 4600 radiomic features were extracted, of which 12 were finally selected. The values of the AUC (area under the subject operating curve) in the training group, the validation group, and overall were 0.911, 0.931, and 0.919, respectively.
This study introduced a novel radiometric and computer algorithm based on DTI images, which can help increase the sensitivity and specificity of PD screening.
基于临床症状和量表评分对帕金森病（PD）进行诊断大多是客观的，而神经影像学用于PD诊断的准确性仍存在争议。本研究旨在引入一种基于扩散张量成像（DTI）指标的放射组学工具，以提高诊断的敏感性和特异性。
在这项基于机器学习的回顾性研究中，我们收集了54名健康对照者（HCs）和56名PD患者的基本临床信息和DTI图像。其中，60名受试者（30名PD患者和30名HCs）被分配到训练组，而测试队列包括26名PD患者和24名HCs。使用新开发的图像处理软件Ray-plus进行特征提取和选择后，采用套索回归确定放射组学特征。
共提取了4600个放射组学特征，最终选择了12个。训练组、验证组和总体的受试者工作曲线下面积（AUC）值分别为0.911、0.931和0.919。
本研究引入了一种基于DTI图像的新型放射测量和计算机算法，有助于提高PD筛查的敏感性和特异性。
