Unstable Prompts, Unreliable Segmentations: A Challenge for Longitudinal Lesion Analysis

2025-07-25

AI Summary

General-purpose lesion segmentation models (e.g., ULS23) suffer from degradation in longitudinal CT analysis due to inter-scan registration errors, causing input spatial misalignment, inconsistent segmentation, and failure in lesion correspondence—revealing the fundamental limitation of single-timepoint modeling. Method: We systematically evaluate ULS23’s temporal consistency and robustness on a public baseline–follow-up CT dataset under controlled spatial displacements. Contribution/Results: Experiments demonstrate that ULS23’s performance is highly sensitive to input localization, exposing the fragility of the lesion-center assumption and underscoring the necessity of joint cross-timepoint modeling. We thus propose a new paradigm: end-to-end temporal integration models explicitly designed for longitudinal analysis. Our findings identify registration error as a critical bottleneck for reliable lesion tracking, providing both theoretical justification and methodological guidance for developing robust longitudinal segmentation frameworks. (149 words)

Technology Category

• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Longitudinal lesion analysis is crucial for oncological care, yet automated tools often struggle with temporal consistency. While universal lesion segmentation models have advanced, they are typically designed for single time points. This paper investigates the performance of the ULS23 segmentation model in a longitudinal context. Using a public clinical dataset of baseline and follow-up CT scans, we evaluated the model's ability to segment and track lesions over time. We identified two critical, interconnected failure modes: a sharp degradation in segmentation quality in follow-up cases due to inter-scan registration errors, and a subsequent breakdown of the lesion correspondence process. To systematically probe this vulnerability, we conducted a controlled experiment where we artificially displaced the input volume relative to the true lesion center. Our results demonstrate that the model's performance is highly dependent on its assumption of a centered lesion; segmentation accuracy collapses when the lesion is sufficiently displaced. These findings reveal a fundamental limitation of applying single-timepoint models to longitudinal data. We conclude that robust oncological tracking requires a paradigm shift away from cascading single-purpose tools towards integrated, end-to-end models inherently designed for temporal analysis.

Problem

• Evaluates ULS23 model's longitudinal lesion segmentation accuracy
• Identifies registration errors causing segmentation quality degradation
• Highlights need for end-to-end models in temporal lesion analysis

Innovation

• Evaluates ULS23 model for longitudinal lesion segmentation
• Identifies registration errors degrading follow-up segmentation quality
• Proposes end-to-end models for robust temporal lesion analysis