Related Experiment Video
Updated: Jun 29, 2026
Guidelines and Experience Using Imaging Biomarker Explorer (IBEX) for Radiomics
Published on: January 8, 2018
MRCP-derived radiomics features associated with biliary dilation: a multi-step feature reduction, multivariable
Alok Verma1, Naresh Kumar Mangalhara1, Shashank Sharma1
1Department of Interventional Radiology, Sawai Man Singh Medical College and Hospital, Jaipur, India.
Magnetic Resonance Cholangiopancreatography (MRCP)-derived radiomics features show promise in identifying biliary dilation. These radiomic features demonstrate strong diagnostic performance, even in borderline cases, suggesting potential for improved clinical assessment.
Area of Science:
- Radiomics and Medical Imaging
- Hepatobiliary and Pancreatic Diseases
- Quantitative Imaging Biomarkers
Background:
- Biliary dilation is a key indicator of various hepatobiliary conditions.
- Accurate identification of biliary dilation is crucial for timely diagnosis and treatment.
- Magnetic Resonance Cholangiopancreatography (MRCP) is a non-invasive imaging modality for evaluating the biliary tree.
Purpose of the Study:
- To identify radiomics features from MRCP images associated with image-defined biliary dilation.
- To assess the independence of these features from demographic confounders.
- To evaluate the diagnostic performance of radiomics in classifying biliary dilation, including challenging borderline cases.
Main Methods:
- Retrospective analysis of 135 patients undergoing MRCP.
- Independent radiologist classification of biliary dilation.
- 3D segmentation of the common bile duct (CBD) and radiomics feature extraction using PyRadiomics.
- Feature selection using reproducibility, variance, correlation, and statistical thresholds.
- Multivariable regression adjusted for demographic factors.
- Classification models (Random Forest, Logistic Regression, Naïve Bayes) trained on texture features.
Main Results:
- Twelve features (4 shape, 8 texture) met selection criteria, with RunLengthNonUniformity and GrayLevelNonUniformity as strongest discriminators.
- Eleven features were significant independent correlates of biliary dilation (p < 0.001).
- Random Forest model achieved high performance in the full cohort (AUC=0.963, accuracy=92.1%) and promising results in the borderline subgroup (AUC=0.832).
Conclusions:
- MRCP-derived radiomics features are significantly associated with image-defined CBD dilation.
- These features demonstrate robust diagnostic performance, including in equivocal borderline cases.
- Prospective multicenter validation is necessary prior to clinical implementation.
More Related Videos
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
05:37A 3D Quantification Technique for Liver Fat Fraction Distribution Analysis Using Dixon Magnetic Resonance Imaging
Published on: October 20, 2023