Hepatocellular adenoma subtyping by qualitative MRI features and machine learning algorithm of integrated qualitative
X Liu1, O Espin-Garcia2, F Khalvati3
1Joint Department of Medical Imaging, University Health Network, Sinai Health System, University of Toronto, 585 University Ave., Toronto, ON, M5G 2N2, Canada.
Clinical Radiology
|June 26, 2023
Summary
Hepatocellular adenoma (HCA) subtyping is improved using qualitative MRI features and machine learning. Quantitative MRI features also aid in diagnosing specific HCA subtypes, offering promising clinical management insights.
Area of Science:
- Radiology
- Oncology
- Medical Imaging
Background:
- Hepatocellular adenoma (HCA) is a rare benign liver tumor with varying subtypes.
- Accurate subtyping of HCA is crucial for clinical management and risk stratification.
- Current subtyping relies heavily on histopathology, which can be invasive.
Purpose of the Study:
- To evaluate the accuracy of qualitative magnetic resonance imaging (MRI) features for HCA subtyping.
- To assess the feasibility of using machine learning (ML) with qualitative and quantitative MRI features for HCA differentiation.
- To compare ML-based subtyping with histopathology as the reference standard.
Main Methods:
- Retrospective analysis of 39 histopathologically subtyped HCAs (HHCA, IHCA, BHCA, UHCA).
- Qualitative MRI features were assessed by two blinded radiologists using a proposed schema.
- Machine learning algorithms (random forest, SVM, logistic regression) were applied to qualitative and quantitative MRI data.
- 1,409 radiomic features were extracted and reduced to 10 principal components for ML analysis.
Main Results:
- Qualitative MRI features achieved diagnostic accuracies of 87% (HHCA), 82% (IHCA), and 74% (UHCA).
- ML based on qualitative MRI features showed AUCs ranging from 0.642 to 0.846 for different subtypes.
- Quantitative radiomic features demonstrated AUCs of 0.83 and 0.82 for predicting HHCA, with 72% sensitivity and 85% specificity.
Conclusions:
- Integrated qualitative MRI features and ML algorithms provide high accuracy for HCA subtyping.
- Quantitative radiomic features are valuable for diagnosing the HHCA subtype.
- These MRI-based approaches show promise for improved clinical management of HCA patients.
