Effective Lymph Nodes Detection in CT Scans Using Location Debiased Query Selection and Contrastive Query Representation in Transformer

2024-04-04

European Conference on Computer Vision

AI Summary

Lymph node (LN) detection in CT scans suffers from low recall and high false-positive rates due to their scattered distribution, low contrast, and morphological similarity to adjacent anatomical structures (e.g., vessels, esophagus). To address these challenges, we propose LN-DETR—a Transformer-based detector featuring: (1) multi-scale 2.5D feature fusion to encode 3D anatomical context; (2) a novel localization-unbiased query selection mechanism to mitigate boundary ambiguity–induced localization errors; (3) query-level contrastive learning to suppress false positives caused by anatomical confounders; and (4) an integrated IoU prediction head with position-deviation–aware query initialization. Evaluated on a large multi-site dataset of 1,067 CT scans, LN-DETR achieves a 4–5% average recall gain at matched false-positive rates. On the NIH DeepLesion benchmark, it attains 88.46% average recall at 0.5–4 false positives per image, establishing new state-of-the-art performance.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Computer Vision: Language and Vision
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

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

Abstract

Lymph node (LN) assessment is a critical, indispensable yet very challenging task in the routine clinical workflow of radiology and oncology. Accurate LN analysis is essential for cancer diagnosis, staging, and treatment planning. Finding scatteredly distributed, low-contrast clinically relevant LNs in 3D CT is difficult even for experienced physicians under high inter-observer variations. Previous automatic LN detection works typically yield limited recall and high false positives (FPs) due to adjacent anatomies with similar image intensities, shapes, or textures (vessels, muscles, esophagus, etc). In this work, we propose a new LN DEtection TRansformer, named LN-DETR, to achieve more accurate performance. By enhancing the 2D backbone with a multi-scale 2.5D feature fusion to incorporate 3D context explicitly, more importantly, we make two main contributions to improve the representation quality of LN queries. 1) Considering that LN boundaries are often unclear, an IoU prediction head and a location debiased query selection are proposed to select LN queries of higher localization accuracy as the decoder query's initialization. 2) To reduce FPs, query contrastive learning is employed to explicitly reinforce LN queries towards their best-matched ground-truth queries over unmatched query predictions. Trained and tested on 3D CT scans of 1067 patients (with 10,000+ labeled LNs) via combining seven LN datasets from different body parts (neck, chest, and abdomen) and pathologies/cancers, our method significantly improves the performance of previous leading methods by>4-5% average recall at the same FP rates in both internal and external testing. We further evaluate on the universal lesion detection task using NIH DeepLesion benchmark, and our method achieves the top performance of 88.46% averaged recall across 0.5 to 4 FPs per image, compared with other leading reported results.

Problem

• Improving lymph node detection accuracy in CT scans
• Reducing false positives in automatic LN detection
• Enhancing 3D context incorporation for better LN analysis

Innovation

• Multi-scale 2.5D feature fusion for 3D context
• Location debiased query selection for accuracy
• Query contrastive learning to reduce false positives