Segmentation variability and radiomics stability for predicting Triple-Negative Breast Cancer subtype using Magnetic Resonance Imaging

2025-04-02

AI Summary

This study investigates the impact of segmentation variability on radiomic prediction of triple-negative breast cancer (TNBC) subtypes from MRI. Using 244 patient MRI scans from the Duke dataset, segmentation perturbations were simulated via manual contouring variations; SHAP-based interpretable feature selection and logistic regression modeling were employed. Results show that segmentation accuracy changes do not significantly degrade model predictive performance. Crucially, feature stability—quantified by intraclass correlation coefficient (ICC), Pearson correlation, and reliability scores—is decoupled from predictive power: highly predictive features need not be stable, and unstable features encoding peritumoral information retain strong discriminative capability. This challenges the prevailing ICC-dominated stability filtering paradigm in radiomics. The study proposes a refined feature selection criterion: prioritizing both predictive efficacy and biological interpretability over mere segmentation robustness—thereby advancing methodological rigor and clinical relevance in radiomic modeling.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Most papers caution against using predictive models for disease stratification based on unselected radiomic features, as these features are affected by contouring variability. Instead, they advocate for the use of the Intraclass Correlation Coefficient (ICC) as a measure of stability for feature selection. However, the direct effect of segmentation variability on the predictive models is rarely studied. This study investigates the impact of segmentation variability on feature stability and predictive performance in radiomics-based prediction of Triple-Negative Breast Cancer (TNBC) subtype using Magnetic Resonance Imaging. A total of 244 images from the Duke dataset were used, with segmentation variability introduced through modifications of manual segmentations. For each mask, explainable radiomic features were selected using the Shapley Additive exPlanations method and used to train logistic regression models. Feature stability across segmentations was assessed via ICC, Pearson's correlation, and reliability scores quantifying the relationship between feature stability and segmentation variability. Results indicate that segmentation accuracy does not significantly impact predictive performance. While incorporating peritumoral information may reduce feature reproducibility, it does not diminish feature predictive capability. Moreover, feature selection in predictive models is not inherently tied to feature stability with respect to segmentation, suggesting that an overreliance on ICC or reliability scores for feature selection might exclude valuable predictive features.

Problem

• Impact of segmentation variability on radiomics feature stability
• Effect of segmentation accuracy on predictive model performance
• Relationship between feature stability and predictive capability in TNBC

Innovation

• Uses ICC for radiomic feature stability
• Investigates segmentation variability impact
• Employs Shapley method for feature selection