Lung-DETR: Deformable Detection Transformer for Sparse Lung Nodule Anomaly Detection

TL;DR

This paper presents Lung-DETR, a novel deformable detection transformer model that improves lung nodule detection in CT scans by addressing data sparsity and similarity issues through innovative preprocessing and model design, achieving state-of-the-art results.

Contribution

Introduction of Lung-DETR, a deformable detection transformer with custom preprocessing and loss functions, tailored for sparse lung nodule detection in CT images, reflecting real-world clinical challenges.

Findings

Achieved 94.2% F1 score on LUNA16 dataset.

Enhanced detection accuracy with a 95.2% recall.

Reduced false positives compared to existing methods.

Abstract

Accurate lung nodule detection for computed tomography (CT) scan imagery is challenging in real-world settings due to the sparse occurrence of nodules and similarity to other anatomical structures. In a typical positive case, nodules may appear in as few as 3% of CT slices, complicating detection. To address this, we reframe the problem as an anomaly detection task, targeting rare nodule occurrences in a predominantly normal dataset. We introduce a novel solution leveraging custom data preprocessing and Deformable Detection Transformer (Deformable- DETR). A 7.5mm Maximum Intensity Projection (MIP) is utilized to combine adjacent lung slices into single images, reducing the slice count and decreasing nodule sparsity. This enhances spatial context, allowing for better differentiation between nodules and other structures such as complex vascular structures and bronchioles. Deformable-DETR is employed to detect nodules, with a custom focal loss function to better handle the imbalanced dataset. Our model achieves state-of-the-art performance on the LUNA16 dataset with an F1 score of 94.2% (95.2% recall, 93.3% precision) on a dataset sparsely populated with lung nodules that is reflective of real-world clinical data.