FissureNet: A Deep Learning Approach For Pulmonary Fissure Detection in CT Images

Abstract

Pulmonary fissure detection in computed tomography (CT) is a critical component for automatic lobar segmentation. The majority of fissure detection methods use feature descriptors that are hand-crafted, low-level, and have local spatial extent. The design of such feature detectors is typically targeted towards normal fissure anatomy, yielding low sensitivity to weak and abnormal fissures that are common in clinical datasets. Furthermore, local features commonly suffer from low specificity, as the complex textures in the lung can be indistinguishable from the fissure when global context is not considered. We propose a supervised discriminative learning framework for simultaneous feature extraction and classification. The proposed framework, called FissureNet, is a coarse-to-fine cascade of two convolutional neural networks. The coarse-to-fine strategy alleviates the challenges associated with training a network to segment a thin structure that represents a small fraction of the image voxels. FissureNet was evaluated on a cohort of 3706 subjects with inspiration and expiration 3DCT scans from the COPDGene clinical trial and a cohort of 20 subjects with 4DCT scans from a lung cancer clinical trial. On both datasets, FissureNet showed superior performance compared to a deep learning approach using the U-Net architecture and a Hessian-based fissure detection method in terms of area under the precisionrecall curve (PR-AUC). The overall PR-AUC for FissureNet, UNet, and Hessian on the COPDGene (lung cancer) dataset was 0.980 (0.966), 0.963 (0.937), and 0.158 (0.182), respectively. On a subset of 30 COPDGene scans, FissureNet was compared to a recently proposed advanced fissure detection method called derivative of sticks (DoS) and showed superior performance with a PR-AUC of 0.991 compared to 0.668 for DoS.

ALTIS: A fast and automatic lung and trachea CT‐image segmentation method

“…Deep neural networks have achieved great results in many fields, including medical imaging . However, we could not find a network that could be optimized from a small training set, as we do for this study, and used to segment the trachea, left lung, and right lung.…”

Artificial Intelligence in COPD: New Venues to Study a Complex Disease

“…AI is transforming and consolidating the upstream set of automated operations necessary to resolve the diseased lung's structural components. Rulebased approaches to segment the lung and the lobes are being replaced by more reliable and precise deep learning methods based on CNN 19,20,21 . Rule-based methods propelled some of the early research applications of CT-based phenotyping 22,23,24,25,26 , however, these approaches lack of generalization and tend to under-or over-segment regions without well-demarcated edges.…”

Relational Modeling for Robust and Efficient Pulmonary Lobe Segmentation in CT Scans

“…The global features were added to the second stage network to provide contextual guidance, while the second stage network was designed to focus on capturing local details at a high resolution. Their framework has also been successfully applied for pulmonary fissure and lung segmentation tasks [18], [19]. In this work, we also employ a two-stage approach, that is trained in an end-to-end fashion.…”

“…The proposed RTSU-Net is robust and produces accurate lobe segmentations even for scans with severe pathology. • We used a multi-resolution framework similar to [15], [18], [19], however, we train both stages in an end-to-end fashion. This gives the RTSU-Net the ability to capture the global object relationships at the full scan level from the first stage network while extracting local details at the second stage simultaneously in the same optimization process.…”