Abstract:The automatic lung nodule detection system can facilitate the early screening of lung cancer and timely medical interventions. However, there still exist multiple nodule candidates produced by initial rough detection in this system, and how to determine authenticity is a key problem. As this work is often challenged by the radiological heterogeneity of the computed tomography scans and the variable sizes of lung nodules, we put forward a multi-resolution convolutional neural network (CNN) to extract features of various levels and resolutions from different depth layers in the network for classification of lung nodule candidates. Through the use of knowledge transfer, the method can be divided into three steps. First, we transfer knowledge from the source CNN model which has been applied to edge detection and improve the model to a new multi-resolution model which is suitable for the image classification task. Then, the knowledge is transformed from source training progress so that all of the side-output branches in the model will be considered in the calculation. Moreover, the loss function and objective equation are improved to be image-wise calculation rather than pixel-wise. Finally, samples production and data enhancement are performed to train and test a classifier tailored for classification of lung nodule candidates. The experimental results on the LUNA16 data set show that our method gets an accuracy of 0.9733, a precision of 0.9673, and an AUC of 0.9954 while being used for lung nodule candidate classification, which is higher than the scores obtained by most of the state-of-the-art approach. In addition, when the test samples with three different sizes of 26* 26, 36* 36, and 48* 48 are used to test the multi-resolution CNN, the accuracy rate of all three experiments exceed 92.81%, which demonstrates that the proposed model is insensitive to input scales.

Multi-scale Pulmonary Nodule Classification with Deep Feature Fusion Via Residual Network

Pulmonary Nodule Detection Based on Multiscale Feature Fusion

A novel benign and malignant classification model for lung nodules based on multi-scale interleaved fusion integrated network

Combining Multi-Scale Feature Fusion with Multi-Attribute Grading, a CNN Model for Benign and Malignant Classification of Pulmonary Nodules.

Multi-Modal Feature Fusion-Based Multi-Branch Classification Network for Pulmonary Nodule Malignancy Suspiciousness Diagnosis

Deep 3D Multi-Scale Dual Path Network for Automatic Lung Nodule Classification

Benign and Malignant Classification of Pulmonary Nodules Based on Multi-Level Feature Fusion

A diagnostic classification of lung nodules using multiple-scale residual network

Fusing Multi-Dimensional Convolution Neural Network for Lung Nodules Classification

Lung nodule classification using deep feature fusion in chest radiography

Diagnostic Classification of Pulmonary Nodules Using a Multi-scale and Multi-input DenseNet.

Predicting Lung Nodule Malignancies by Combining Deep Convolutional Neural Network and Handcrafted Features

Multi-Level Cross Residual Network for Lung Nodule Classification

3D Spatial Pyramid Dilated Network for Pulmonary Nodule Classification.

Pulmonary nodule classification with deep residual networks

Multi-scale Convolutional Neural Networks for Lung Nodule Classification.

DAG‐Net: Dual‐Branch Attention‐Guided Network for Multi‐Scale Information Fusion in Lung Nodule Segmentation

Classification of Benign and Malignant Lung Nodules Based on Deep Convolutional Network Feature Extraction

Classification of Benign and Malignant Pulmonary Nodules Based on Deep Learning

Research of Lung Cancer-Assisted Diagnosis Algorithm Based on Multi-scale Convolution Kernel Network

Multi-Resolution CNN and Knowledge Transfer for Candidate Classification in Lung Nodule Detection