English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

This paper focuses on a novel approach for false-positive reduction (FPR) ofnodule candidates in Computer-aided detection (CADe) systems following thesuspicious lesions detection stage. Contrary to typical decisions in medicalimage analysis, the proposed approach considers input data not as a 2D or 3Dimage, but rather as a point cloud, and uses deep learning models for pointclouds. We discovered that point cloud models require less memory and arefaster both in training and inference compared to traditional CNN 3D, theyachieve better performance and do not impose restrictions on the size of theinput image, i.e. no restrictions on the size of the nodule candidate. Wepropose an algorithm for transforming 3D CT scan data to point cloud. In somecases, the volume of the nodule candidate can be much smaller than thesurrounding context, for example, in the case of subpleural localization of thenodule. Therefore, we developed an algorithm for sampling points from a pointcloud constructed from a 3D image of the candidate region. The algorithm isable to guarantee the capture of both context and candidate information as partof the point cloud of the nodule candidate. We designed and set up anexperiment in creating a dataset from an open LIDC-IDRI database for a featureof the FPR task, and is herein described in detail. Data augmentation wasapplied both to avoid overfitting and as an upsampling method. Experiments wereconducted with PointNet, PointNet++, and DGCNN. We show that the proposedapproach outperforms baseline CNN 3D models and resulted in 85.98 FROC versus77.26 FROC for baseline models. We compare our algorithm with published SOTAand demonstrate that even without significant modifications it works at theappropriate performance level on LUNA2016 and shows SOTA on LIDC-IDRI.

Deep Learning on Point Clouds for False Positive Reduction at Nodule Detection in Chest CT Scans