DynoDINO: Harnessing Dynamic Latent Information from DINO Features for Multi-Phase Medical Image Segmentation

2026-07-31

AI Summary

This study addresses the challenge of inaccurate lesion segmentation in multiphase contrast-enhanced CT images, which is often compromised by misalignment and temporal discontinuity due to respiratory motion, anatomical inconsistencies, or missing phases. To tackle this issue, the authors propose a unified multiphase segmentation framework that first establishes inter-phase anatomical correspondence through slice-level alignment. They then introduce an adaptive gating mechanism integrating Mix-attention and differential residual learning to effectively model cross-phase temporal dependencies. This design preserves diagnostically critical contrast dynamics while suppressing misalignment artifacts. Evaluated on three large-scale datasets—LiTS, PLC-CECT, and WAW-TACE—the method significantly improves lesion boundary accuracy and structural fidelity under standard, shifted, and missing phase conditions.

Abstract

Multi-phase Contrast-Enhanced Computed Tomography (CECT) plays a central role in the diagnosis and characterization of focal lesions by capturing temporal enhancement patterns across multiple acquisition phases. Accurate lesion segmentation from such data remains challenging because clinically relevant contrast kinetics are distributed across phases, while anatomical inconsistencies, respiratory motion, and incomplete acquisitions often lead to inter-phase misalignment and interrupted temporal information. Conventional segmentation frameworks typically process each phase independently or rely on simple fusion strategies, limiting their temporal reasoning capability. To address these challenges, we propose DynoDINO, a unified framework tailored to address the core challenges of multi-phase medical image segmentation. DynoDINO first performs slice-level alignment to establish inter-phase anatomical correspondence and then employs a Multi-phase Fusion Model to jointly enhance temporal correlations across phases. Our fusion model incorporates a Mix-attention (MA) mechanism for efficient multi-phase feature calibration and an Adaptive Gating Mechanism with difference-based residual learning to selectively preserve diagnostically relevant contrast variations while suppressing artifacts caused by residual misalignment. In addition, the adaptive gating mechanism improves training stability by preventing feature degradation caused by unguided subtraction operations. Experiments on three large-scale datasets, including LiTS, PLC-CECT, and WAW-TACE, demonstrate that DynoDINO consistently improves boundary delineation and structural fidelity under standard, shifted, and missing-phase conditions.

Problem

• multi-phase segmentation
• temporal misalignment
• contrast kinetics
• medical image segmentation
• inter-phase inconsistency

Innovation

• multi-phase segmentation
• feature fusion
• adaptive gating
• temporal correlation
• medical image alignment