Uncertainty quantification for White Matter Hyperintensity segmentation detects silent failures and improves automated Fazekas quantification

2024-11-26

arXiv.org

AI Summary

White matter hyperintensity (WMH) segmentation suffers from “silent failures”—particularly missed small deep WMHs—due to morphological variability, ill-defined boundaries, and intensity similarity with acute infarcts or artifacts. Method: We propose a novel Fazekas scoring framework integrating uncertainty quantification (UQ) with spatial anatomical priors (deep vs. periventricular). For the first time, UQ maps are jointly modeled with anatomical location to actively detect silent failures; we empirically demonstrate that UQ effectively discriminates WMHs from acute infarcts. Segmentation robustness is enhanced via stochastic network architectures and deep ensembles. Evaluation employs Dice score and absolute volume difference (AVD). Results: The framework achieves balanced Fazekas classification accuracy of 0.71 for deep WMHs and 0.82 for periventricular WMHs. Model calibration and abnormal segmentation detection capability are significantly improved, enabling reliable clinical interpretation and error-aware WMH assessment.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Computer Vision: Segmentation

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

White Matter Hyperintensities (WMH) are key neuroradiological markers of small vessel disease present in brain MRI. Assessment of WMH is important in research and clinics. However, WMH are challenging to segment due to their high variability in shape, location, size, poorly defined borders, and similar intensity profile to other pathologies (e.g stroke lesions) and artefacts (e.g head motion). In this work, we apply the most effective techniques for uncertainty quantification (UQ) in segmentation to the WMH segmentation task across multiple test-time data distributions. We find a combination of Stochastic Segmentation Networks with Deep Ensembles yields the highest Dice and lowest Absolute Volume Difference % (AVD) score on in-domain and out-of-distribution data. We demonstrate the downstream utility of UQ, proposing a novel method for classification of the clinical Fazekas score using spatial features extracted for WMH segmentation and UQ maps. We show that incorporating WMH uncertainty information improves Fazekas classification performance and calibration, with median class balanced accuracy for classification models with (UQ and spatial WMH features)/(spatial WMH features)/(WMH volume only) of 0.71/0.66/0.60 in the Deep WMH and 0.82/0.77/0.73 in the Periventricular WMH regions respectively. We demonstrate that stochastic UQ techniques with high sample diversity can improve the detection of poor quality segmentations. Finally, we qualitatively analyse the semantic information captured by UQ techniques and demonstrate that uncertainty can highlight areas where there is ambiguity between WMH and stroke lesions, while identifying clusters of small WMH in deep white matter unsegmented by the model.

Problem

• Quantify uncertainty in White Matter Hyperintensity segmentation to detect failures
• Improve Fazekas score classification using spatial WMH and uncertainty features
• Enhance segmentation accuracy by distinguishing WMH from stroke lesions and artifacts

Innovation

• UQ techniques reduce silent segmentation failures
• Stochastic Networks with Deep Ensembles improve accuracy
• Spatial features from UQ maps enhance Fazekas scoring