Multi-Branch Ensemble Learning Architecture Based on 3D CNN for False Positive Reduction in Lung Nodule Detection

Abstract

It is critical to have accurate detection of lung nodules in CT images for the early diagnosis of lung cancer. In order to achieve this, it is necessary to reduce the false positive rate of detection. Due to the heterogeneity of lung nodules and their similarity to the background, it is difficult to distinguish true lung nodules from numerous candidate nodules. In this paper, in order to solve this challenging problem, we propose a Multi-Branch Ensemble Learning architecture based on the three-dimensional (3D) convolutional neural networks (MBEL-3D-CNN). The method combines three key ideas: 1) constructing a 3D-CNN to make the maximum utilization of spatial information of lung nodules in the 3D space; 2) embedding a multi-branch network architecture in the 3D-CNN that is well adapted to the heterogeneity of lung nodules, and; 3) using ensemble learning to effectively improve the generalization performance of the 3D-CNN model. In addition, we use offline hard mining operations to make the network capable of handling those indistinguishable positive and negative samples. The proposed method was tested on the dataset LUNA16 in our experiments. The experimental results show that MBEL-3D-CNN architecture can achieve better screening results.INDEX TERMS Computer-aided diagnosis, ensemble learning, false positive reduction, offline hard mining, 3D CNN.

Pulmonary Nodule Detection Using 3-D Residual U-Net Oriented Context-Guided Attention and Multi-Branch Classification Network

“…Method [10] designed the 3-D input feature maps of three sizes, and the fusion weight was manually determined to classify pulmonary nodules, but its detection rates needed further improvement. Method [28] integrated with three 3-D CNN subnetworks to classify pulmonary nodules at the expense of a high network complexity. Method [11] designed a 3-D input feature map and multi-scale feature layers were used for feature fusion to improve the feature extraction for high classification accuracy, but the microscopic information of nodules is not fully utilized.…”

“…Hamidian et al [27] used a 3-D fully convolutional network (FCN) to generate a score map for nodule candidate identification, and employed another 3-D CNN for nodule and non-nodule discrimination. Therefore, Cao et al [28] proposed a three-dimensional convolutional neural network based on a multi-branch set learning algorithm for pulmonary nodule detection, three branching structures based on VggNet, IResNet and DenseNet are constructed to correspond to three input sizes. In order to make full use of 3-D spatial context feature information, 3-D CNN was used in these branching structures to extract different deep features of pulmonary nodules, and then three classification results were integrated together to improve the detection results.…”

Computer-aided diagnostic system kinds and pulmonary nodule detection efficacy

“…After that, these structures are merged and made a feature extractor system. In [67], 3-D CNN is adopted with multi-level contextual. Due to variation in the size of nodules, so four various size of 3-D CNN is designed and the fusion for these networks give an excellent coverage error for classifiers.…”

A novel lung nodule accurate detection of computerized tomography images based on convolutional neural network and probability graph model

“…We instead use a class-balancing weight 𝛽 to automatically balance the label between nodule/normal classes. In Equation (10), we define the cross-entropy loss function of class-balanced in each slice image.…”