End-to-end lung nodule detection framework with model-based feature projection block
This addresses the problem of accurate and efficient lung nodule detection for medical imaging, with incremental improvements in sensitivity and false-positive reduction.
The paper tackles lung nodule detection in chest CT scans by proposing an end-to-end framework with a model-based feature projection block, achieving state-of-the-art results on LUNA2016 with 0.959 average sensitivity and 0.936 sensitivity at a false-positive level of 0.25 per scan.
This paper proposes novel end-to-end framework for detecting suspicious pulmonary nodules in chest CT scans. The method core idea is a new nodule segmentation architecture with a model-based feature projection block on three-dimensional convolutions. This block acts as a preliminary feature extractor for a two-dimensional U-Net-like convolutional network. Using the proposed approach along with an axial, coronal, and sagittal projection analysis makes it possible to abandon the widely used false positives reduction step. The proposed method achieves SOTA on LUNA2016 with 0.959 average sensitivity, and 0.936 sensitivity if the false-positive level per scan is 0.25. The paper describes the proposed approach and represents the experimental results on LUNA2016 as well as ablation studies.