Concept guided model improves lung lesion segmentation accuracy
CGSM: Concept-Guided Segmentation Model for Precise Pulmonary Lesion Delineation
Computer Vision and Pattern Recognition
Summary
Identifying lung lesions accurately in medical images is important for diagnosis and treatment, but existing methods often miss fine details or struggle with small lesions. The authors propose a new approach called CGSM that uses concepts generated by large language models and reviewed by experts to guide the segmentation process. This method combines text and image information more effectively to outline lesions precisely. Tests on public datasets show CGSM performs better than previous models in detecting lung lesions.
What this means in practice
- •For medical imaging teams: Improve automated lung lesion boundary detection in CT scans with integration of concept-guided segmentation for more precise diagnostics.
- •For radiology software developers: Implement concept-enhanced segmentation modules in clinical software to better delineate small-scale pulmonary lesions from imaging data.
Authors
Changheng Lin, Wenjie Zhang, Yushan Lu, Xinyue Yan, Xiao Jia, Wei Zhang
Abstract
Accurate segmentation of pulmonary lesions is essential for effective clinical diagnosis and treatment strategies. Existing segmentation approaches often lack task-specific semantic guidance, as text-based annotations typically offer coarse localization of lesions, leading to inadequate delineation of lesion boundaries and poor performance on small-scale lesions. To address this, we propose CGSM, a Concept-Guided Segmentation Model that integrates LLM-generated and clinically reviewed concepts into the segmentation process. Specifically, we design a Concept-Visual Alignment Module (CVAM) to activate relevant tokens within the concepts that align with visual features, enhancing the interaction between textual and visual information. In addition, we introduce a Concept Modulated Decoder (CM-Decoder), which uses concepts from CVAM as modulation signals to facilitate the adaptive fusion of image and text features, improving the segmentation accuracy. Extensive experiments on two public datasets show that CGSM achieves state-of-the-art performance, with results of 91.59% Dice and 84.49% mIoU on the QaTa-COV19 dataset, demonstrating its effectiveness in pulmonary lesion segmentation.