%0 Journal Article
%T 合成磁共振影像组学模型对前列腺良恶性病变的鉴别诊断价值<br>Differential Value of Synthetic MRI Radiomics Model for Benign and Malignant Prostate Lesions
%A 黄龙龙
%A 徐文娟
%A 杜芳
%J Medical Diagnosis
%P 98-105
%@ 2164-5418
%D 2024
%I Hans Publishing
%R 10.12677/MD.2024.141014
%X 目的：探讨基于合成磁共振(SyMRI)的影像组学模型对前列腺良恶性病变的鉴别诊断价值。方法：回顾性分析本院2020年11月~2022年10月经病理证实的213例患者的影像资料及临床资料，其中良性108例，恶性105例。患者均行常规MRI和多动态多回波(MDME)序列扫描。采用随机抽样方法将患者按7:3的比例分成训练组与验证组。利用Python 3.7.1对患者的影像资料进行特征采集，采用T检验/Mann-Whitney U检验、mRMR、LASSO算法及Spearman相关分析筛选影像组学特征，使用逻辑回归(LR)、支持向量机(SVM)和随机森林(RF)构建T1、T2、PD及多序列定量图的影像组学模型。通过受试者工作特征曲线(ROC)对模型的诊断效能进行验证。结果：在不同定量图的影像组学模型中，多序列联合模型的效能最好，LR、RF及SVM的ROC曲线下面积(AUC)训练组分别为0.859、0.905、0.852，验证组分别为0.850、0.889、0.847，其中RF模型的AUC值最高。结论：合成磁共振影像组学模型对前列腺良恶性病变有较高的鉴别诊断效能，其中多序列联合模型的诊断效能最好。<br />
Objective: To explore the differential value of synthetic magnetic resonance (SyMRI) based radi-omics model for benign and malignant prostate lesions. Methods: The imaging and clinical data of 213 patients with pathologically confirmed prostate lesions from November 2020 to October 2022 in our hospital were retrospectively analyzed, of which 108 were benign and 105 were malignant. Patients were scanned with conventional MRI and multidynamic multiple echo (MDME) sequences. Random sampling method was used to divide the patients into training and validation groups in the ratio of 7:3. Python 3.7.1 was used to collect features from the patients’ images, and T-test/Mann-Whitney U-test, mRMR, LASSO algorithm and Spearman’s correlation analysis were used to screen the imaging histological features, and logistic regression (LR), support vector ma-chine (SVM) and random forest (RF) were used to construct the T1, T2, PD and multi-sequence quantile map. Quantile map and multiple sequence quantile map for imaging histology modeling. The diagnostic efficacy of the models was validated by subject work characteristic curve (ROC). Re-sults: Radiomics model in different quantitative maps, the combined multisequence model had the best efficacy, with area under the ROC curve (AUC) of 0.859, 0.905, and 0.852 for LR, RF, and SVM in the training group and 0.850, 0.889, and 0.847 in the validation group, respectively, with the RF model having the highest AUC value. Conclusion: The synthetic magnetic resonance imaging histol-ogy model has high differential diagnostic efficacy for benign and malignant prostate lesions, with the best diagnostic efficacy of the combined multi-sequence model.
%K 合成磁共振，影像组学，前列腺癌，诊断效能<br>SyMRI
%K Radiomics
%K Prostate Cancer
%K Diagnostic Efficacy
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=82672