Abstract:<section class="article-section__content"><h3 class="article-section__sub-title section1"> Pupose</h3><p>The research is to improve the efficiency and accuracy of recognition of honeycomb lung in CT images.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Methods</h3><p>Deep learning methods are used to achieve automatic recognition of honeycomb lung in CT images, however, is time consuming and less accurate due to the large amount of structural parameters. In this paper, a novel recognition method based on MobileNetV1 network, multi‐scale feature fusion method (MSFF) and dilated convolution are explored to deal with honeycomb lung in CT images classification. Extract features in dilated convolution of different dilated rate to expand the receptive field and keep the features not lost, and fuse the features of different scales at multi‐scale feature fusion block are the first adopted to solve the problem of feature loss and incomplete feature extraction. After that, by using linear activation functions (Sigmoid) instead of nonlinear activation functions (ReLu) in the improved deep separable convolution blocks to retain the feature information of each channels. Finally, by reducing the number of improved deep separable blocks to reduce the computation and resource consumption of the model.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Results</h3><p>The experimental results show that improved MobileNet model has the best performance and the potential for recognition on honeycomb lung image datasets, which includes 6318 images. By comparing with 4 traditional models (SVM, RF, decision‐tree and KNN) and 11 deep learning models (LeNet‐5, AlexNet, VGG‐16, GoogleNet, ResNet18, DenseNet121, SENet18, InceptionV3, InceptionV4, Xception, MobileNetV1). Our model achieved the performance with an accuracy of 99.52%, a sensitivity of 99.35%, a specificity of 99.89%.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Conclusion</h3><p>Improved MobileNet model is designed for the automatic recognition and classification of honeycomb lung in CT images. Through experiments comparative analysis for other models of machine learning and deep learning, it is proved that the proposed improved MobileNet method has the best recognition accuracy with fewer the model parameters and less the calculation time.</p></section>

SM-GRSNet: sparse mapping-based graph representation segmentation network for honeycomb lung lesion

An Efficient Honeycomb Lung Segmentation Network Combining Multi-Paradigms Representation and Cascade Attention

MCAFNet: multiscale cross-layer attention fusion network for honeycomb lung lesion segmentation

SM-RNet: A Scale-Aware-Based Multiattention-Guided Reverse Network for Pulmonary Nodules Segmentation.

Multitask Learning for Concurrent Grading Diagnosis and Semi-Supervised Segmentation of Honeycomb Lung in CT Images

MCSC-UTNet: Honeycomb Lung Segmentation Algorithm Based on Separable Vision Transformer and Context Feature Fusion.

Segmentation Algorithm for Honeycomb Lung CT Images Based on Pyramidal Pooling Channel Transformer

H-SegNet: hybrid segmentation network for lung segmentation in chest radiographs using mask region-based convolutional neural network and adaptive closed polyline searching method

RPLS-Net: Pulmonary Lobe Segmentation Based on 3D Fully Convolutional Networks and Multi-Task Learning

Multiscale lung nodule segmentation based on 3D coordinate attention and edge enhancement

Anterior Mediastinal Lesion Segmentation Based on Two-Stage 3D ResUNet With Attention Gates and Lung Segmentation

Accurate Lung Nodules Segmentation with Detailed Representation Transfer and Soft Mask Supervision

Recognition of honeycomb lung in CT images based on improved MobileNet model

Uncertainty-guided cross learning via CNN and transformer for semi-supervised honeycomb lung lesion segmentation

Multi-granularity scale-aware networks for hard pixels segmentation of pulmonary nodules

SCLMnet: A dual-branch guided network for lung and lung lobe segmentation

Texture Feature-Aware Consistency for Semi-Supervised Honeycomb Lung Lesion Segmentation

Weakly-Supervised Segmentation-Based Quantitative Characterization of Pulmonary Cavity Lesions in CT Scans

SGNet: Structure-Aware Graph-Based Network for Airway Semantic Segmentation

Multi-Attention Segmentation Networks Combined with the Sobel Operator for Medical Images

A deep learning-based post-processing method for automated pulmonary lobe and airway trees segmentation using chest CT images in PET/CT