GRASPing Anatomy to Improve Pathology Segmentation

2025-08-05

AI Summary

Existing deep learning methods for pathological segmentation neglect anatomical context, limiting segmentation accuracy. To address this, we propose GRASP—a modular, plug-and-play framework that injects anatomical prior knowledge into pathological segmentation without retraining anatomical models. GRASP employs a dual-path anatomical information integration mechanism: (1) leveraging anatomical structure pseudo-labels as auxiliary inputs, and (2) introducing a Transformer-guided feature alignment and multimodal fusion module to jointly optimize anatomical and pathological representations in PET/CT images. The framework is architecture-agnostic and requires no architectural modification of the base segmentation model. Evaluated on two clinical PET/CT datasets, GRASP consistently achieves state-of-the-art performance across multiple metrics. Ablation studies confirm the efficacy of each component, demonstrating significant and consistent improvements in both accuracy and robustness for diverse segmentation architectures.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Radiologists rely on anatomical understanding to accurately delineate pathologies, yet most current deep learning approaches use pure pattern recognition and ignore the anatomical context in which pathologies develop. To narrow this gap, we introduce GRASP (Guided Representation Alignment for the Segmentation of Pathologies), a modular plug-and-play framework that enhances pathology segmentation models by leveraging existing anatomy segmentation models through pseudolabel integration and feature alignment. Unlike previous approaches that obtain anatomical knowledge via auxiliary training, GRASP integrates into standard pathology optimization regimes without retraining anatomical components. We evaluate GRASP on two PET/CT datasets, conduct systematic ablation studies, and investigate the framework's inner workings. We find that GRASP consistently achieves top rankings across multiple evaluation metrics and diverse architectures. The framework's dual anatomy injection strategy, combining anatomical pseudo-labels as input channels with transformer-guided anatomical feature fusion, effectively incorporates anatomical context.

Problem

• Enhance pathology segmentation using anatomical context
• Integrate anatomy models without retraining auxiliary components
• Improve accuracy via pseudo-labels and feature alignment

Innovation

• Modular plug-and-play framework GRASP
• Pseudolabel integration and feature alignment
• Dual anatomy injection strategy