Feasibility of Radiomics Combined with Brisque for Automated Assessment of Hepatic Gadolinium Ethoxybenzyl Diethylenetriamine Pentaacetic Acid Dynamic Enhanced Arterial Phase Image Quality
10.3969/j.issn.1005-5185.2025.06.015
- VernacularTitle:影像组学联合Brisque自动评估肝脏钆塞酸二钠动态增强动脉期图像质量的可行性
- Author:
Tingting LI
1
;
Xiaowen WANG
1
;
Ying HUANG
1
;
Lan ZHANG
1
Author Information
1. 河南中医药大学第一附属医院MRI科,河南中医药大学中医药信息智能分析与利用郑州市重点实验室,河南 郑州 450000
- Publication Type:Journal Article
- Keywords:
Magnetic resonance imaging;
Liver;
Radiomics;
Brisque;
Gd-EOB-DTPA;
No reference image quality assessment
- From:
Chinese Journal of Medical Imaging
2025;33(6):651-656
- CountryChina
- Language:Chinese
-
Abstract:
Purpose To explore the feasibility and validity of radiomics combined with blind/referenceless image spatial quality evaluator(Brisque)for automated assessment of arterial-phase image quality in hepatic gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid(Gd-EOB-DTPA)dynamic enhanced MRI examinations.Materials and Methods A total of 326 arterial phase images from January 2020 to March 2024 in the First Affiliated Hospital of Henan University of Chinese Medicine for liver Gd-EOB-DTPA dynamic enhancement MRI examination were retrospectively collected.A 4-level(1-4 points)Likert scale was used for subjective image quality scoring,and then binary classification was used to divide the images into clinically acceptable group(166 images with a score of 2-4)and clinically unacceptable group(160 images with a score of 1).All images were randomly divided into a training set(227 images)and a test set(99 images)in a 7∶3 ratio.Brisque images were created based on arterial-phase images,and subsequently Brisque image features,image histologic features of the original arterial-phase images and combined features of the two were extracted.The extracted data were reduced in dimension and the optimal features of radiomics were selected.Based on the Logistic regression algorithm,radiomics models,Brisque models and combined models of the two were constructed.The receiver operating characteristic curve was drawn for each subject,and the area under the curve(AUC),threshold,sensitivity,specificity,accuracy,negative predictive value and positive predictive value were calculated to analyze the diagnostic performance of each model,and the net benefit of the models was evaluated by decision curve analysis.Results The AUC values of the radiomics model,Brisque model and combined model in the training set were 0.942(95%CI 0.914-0.969),0.722(95%CI 0.655-0.788)and 0.967(95%CI 0.947-0.987),respectively.In the test set,the AUC values were 0.904(95%CI 0.848-0.960),0.684(95%CI 0.579-0.789)and 0.920(95%CI 0.866-0.972),respectively.There was statistically significant difference between two of the three models in the training set(Z=6.493,3.397,7.427;all P<0.05);in the test set,there was statistically significant difference between the combined model and the Brisque model,and between the radiomics model and the Brisque model(Z=4.137,3.740,both P<0.05),but no statistically significant difference between the combined model and the radiomics model(P>0.05).The diagnostic performance was higher in the combined model in both the training set and the test set.The decision curve analysis curve suggested that the net benefit of the combined model was greater than that of the other two models.Conclusion Radiomics combined with Brisque can be applied in the assessment of the quality of arterial-phase images in Gd-EOB-DTPA dynamic-enhanced MRI examinations of the liver to automatically detect clinically unacceptable arterial-phase images of the liver with high diagnostic efficacy,which contributes to the improvement of the quality of the images at the time of acquisition,leading to a reduction in the number of duplicate scans and an increase in the accuracy of the diagnosis.
