Multi-scale Dilated Convolution Research Articles

Cross-domain meta learning fault diagnosis based on multi-scale dilated convolution and adaptive relation module

For the established fault classification system, new faults cannot be identified due to lack of training data in the process of equipment operation. Aiming at the problems of multi-classification, small samples, and cross-domain brought by the new faults, one meta learning intelligent fault diagnosis method is proposed based on multi-scale dilated convolution and relation module. Firstly, multi-scale convolution is utilized to improve the feature extraction effectiveness in the extraction module. Subsequently, the fusion module is designed by dilated convolution and stochastic pooling. Finally, the relation module is employed to evaluate the distance between samples for fault diagnosis. Crucially, the meta learning strategy is executed to transform the training set into multiple tasks to train the proposed method. The superiority and effectiveness of the proposed method is validated by bearing and gearbox experiments with a few labeled fault samples. For the bearing fault diagnosis test, the verification results show that the accuracy rate of this method is 95.11% in 8way 1-shot, which is increased by 6.15% on average.

A novel computer-assisted diagnosis method of knee osteoarthritis based on multivariate information and deep learning model

Knee Osteoarthritis (KOA) is a chronic joint disease characterized by degeneration of knee's articular cartilage. Comparing with the imaging diagnostic method, the vibroarthrographic (VAG) may be emerged as a new candidate for diagnosis of KOA in clinics. However, it is challenging for doctors to evaluate patients' condition by visually detecting VAGs due to the limited understanding about contained information. Besides, the basic physiological signals are also closely correlated with a greater risk of KOA. Based on this, we focus on studying computer-assisted diagnosis method of KOA (KOA-CAD) using multivariate information (i.e. VAGs and basic physiological signals) based on an improved deep learning model (DLM). Firstly, a new Laplace distribution-based strategy (LD-S) for classification in DLM is designed. Secondly, an aggregated multiscale dilated convolution network (AMD-CNN) is constructed to learn features from multivariate information of KOA patients. Then, a new KOA-CAD method is proposed by integrating the AMD-CNN with the LD-S to realize three CAD objectives, including the automatic KOA detection, the KOA early detection, and the KOA grading detection. Finally, the multivariate information collected clinically is applied to verify the proposed method, where the automatic detection accuracy, early detection accuracy and grading detection accuracy are 93.6%, 92.1%, and 84.2% respectively.

Superresolution Reconstruction Algorithm of Ultrasonic Logging Images Based on High-Frequency Enhancement

High-resolution logging images with glaring detail information are useful for analysing geological features in the field of ultrasonic logging. The resolution of logging images is, however, severely constrained by the complexity of the borehole and the frequency restriction of the ultrasonic transducer. In order to improve the image superresolution reconstruction algorithm, this paper proposes a type of ultrasonic logging based on high-frequency characteristics, with multiscale dilated convolution to feature as the basis of network-learning blocks, training in the fusion of different scale texture feature. The outcomes of other superresolution reconstruction algorithms are then compared to the outcomes of the two-, four-, and eightfold reconstruction. The proposed algorithm enhances subjective vision while also enhancing PSNR and SSIM evaluation indexes, according to a large number of experiments.

Automatic segmentation method using FCN with multi-scale dilated convolution for medical ultrasound image

Automatic segmentation method using FCN with multi-scale dilated convolution for medical ultrasound image

Do you need sharpened details? Asking MMDC-Net: Multi-layer multi-scale dilated convolution network for retinal vessel segmentation

Convolutional neural networks (CNN), especially numerous U-shaped models, have achieved great progress in retinal vessel segmentation. However, a great quantity of global information in fundus images has not been fully explored. And the class imbalance problem of background and blood vessels is still serious. To alleviate these issues, we design a novel multi-layer multi-scale dilated convolution network (MMDC-Net) based on U-Net. We propose an MMDC module to capture sufficient global information under diverse receptive fields through a cascaded mode. Then, we place a new multi-layer fusion (MLF) module behind the decoder, which can not only fuse complementary features but filter noisy information. This enables MMDC-Net to capture the blood vessel details after continuous up-sampling. Finally, we employ a recall loss to resolve the class imbalance problem. Extensive experiments have been done on diverse fundus color image datasets, including STARE, CHASEDB1, DRIVE, and HRF. HRF has a large resolution of 3504 × 2336 whereas others have a small resolution of slightly more than 512 × 512. Qualitative and quantitative results verify the superiority of MMDC-Net. Notably, satisfactory accuracy and sensitivity are acquired by our model. Hence, some key blood vessel details are sharpened. In addition, a large number of further validations and discussions prove the effectiveness and generalization of the proposed MMDC-Net.

Segmentation and volume quantification of epicardial adipose tissue in computed tomography images.

Many cardiovascular diseases are closely related to the composition of epicardial adipose tissue (EAT). Accurate segmentation of EAT can provide a reliable reference for doctors to diagnose the disease. The distribution and composition of EAT often have significant individual differences, and the traditional segmentation methods are not effective. In recent years, deep learning method has been gradually introduced into EAT segmentation task. The existing EAT segmentation methods based on deep learning have a large amount of computation and the segmentation accuracy needs to be improved. Therefore, the purpose of this paper is to develop a lightweight EAT segmentation network, which can obtain higher segmentation accuracy with less computation and further alleviate the problem of false-positive segmentation. First, the obtained computed tomography was preprocessed. That is, the threshold range of EAT was determined to be -190, -30HU according to prior knowledge, and the non-adipose pixels were excluded by threshold segmentation to reduce the difficulty of training. Second, the image obtained after thresholding was input into the lightweight RDU-Net network to perform the training, validating, and testing process. RDU-Net uses a residual multi-scale dilated convolution block in order to extract a wider range of information without changing the current resolution. At the same time, the form of residual connection is adopted to avoid the problem of gradient expansion or gradient explosion caused by too deep network, which also makes the learning easier. In order to optimize the training process, this paper proposes PNDiceLoss, which takes both positive and negative pixels as learning targets, fully considers the class imbalance problem, and appropriately highlights the status of positive pixels. In this paper, 50 CCTA images were randomly selected from the hospital, and the commonly used Dice similarity coefficient (DSC), Jaccard similarity, accuracy (ACC), specificity (SP), precision (PC), and Pearson correlation coefficient are used as evaluation metrics. Bland-Altman analysis results show that the extracted EAT volume is consistent with the actual volume. Compared with the existing methods, the segmentation results show that the proposed method achieves better performance on these metrics, achieving the DSC of 0.9262. The number of false-positive pixels has been reduced by more than half. Pearson correlation coefficient reached 0.992, and linear regression coefficient reached 0.977 when measuring the volume of EAT obtained. In order to verify the effectiveness of the proposed method, experiments are carried out in the cardiac fat database of VisualLab. On this database, the proposed method also achieved good results, and the DSC value reached 0.927 in the case of only 878 slices. A new method to segment and quantify EAT is proposed. Comprehensive experiments show that compared with some classical segmentation algorithms, the proposed method has the advantages of shorter time-consuming, less memory required for operations, and higher segmentation accuracy. The code is available at https://github.com/lvanlee/EAT_Seg/tree/main/EAT_seg.

Multi-scale dilated convolutional network for knowledge graph embedding

It aims to learn how to represent the low-dimensional vectors of entities and relations using observed triplets. KGE can benefit a variety of downstream tasks, such as KG completion and triplet classification. Deep models achieve state-of-the-art performance by leveraging the powerful nonlinear fitting ability of neural networks. However, most existing methods ignore multi-scale interaction features between entities and relations except InceptionE, which is hard to train because of high computation costs. In this paper, we propose a new KGE model called MDCE, that uses multi-scale dilated convolution to capture rich interaction features at different scales. Meanwhile, MDCE has lower computation costs than InceptionE. We perform extensive experiments on multiple benchmark datasets; results on the link prediction task show that the proposed model MDCE not only significantly outperforms existing state-of-the-art models but is also efficient and robust.

Method of feature pyramid and attention enhancement network for pavement crack detection

The detection of pavement crack information plays an important role in the service evaluation of infrastructure engineering performance. The randomness of road noise, pollution, and spots make it difficult to ensure the accuracy of traditional image processing algorithms to detect cracks. Therefore, the crack identification technology under complex background conditions is particularly important. We propose a network architecture, named feature pyramid and attention enhancement network (FPAEN) for pavement crack detection. Through a new multiscale feature fusion module, the deep semantic information is integrated into the low-level convolution stage layer by layer, so the context information circulates throughout the whole network. In addition, the attention enhancement mechanism and the multiscale dilated convolution module are used to strengthen the network model’s ability to locate the crack pixels. To evaluate the performance of the proposed approach, we train FPAEN on the DeepCrack dataset and test it on four datasets. Experimental results show that our method outperforms other advanced crack detection approaches in F-score and mean intersection over union.

Enhancing ensemble diversity based on multiscale dilated convolution in image classification

Convolutional neural networks (CNNs) have achieved extraordinary success on many image classification tasks in recent years. The use of dilated convolution in a CNN can increase the network’s receptive field and improve its performance, and dilated convolution can also be used to compress a CNN to realize a lightweight model. In previous studies, multiscale dilated convolution has been adopted with a focus on improving the internal network structure of a specific CNN model. Because they enable the direct use of pretrained models, transfer learning CNNs (TL-CNNs) have been widely applied for image recognition based on small datasets. This paper proposes a novel multiscale dilated-convolution-based ensemble learning (MDCEL) method for effectively improving the performance of a pretrained CNN model. The authors' primary assumption is that semantic representations of different images can be obtained based on multiscale dilated convolution. Therefore, constructing an ensemble of diverse TL-CNN classifiers makes it possible to achieve higher performance than that offered by the traditional TL-CNN methods. The MDCEL method is highly versatile and can be applied to various conventional pretrained CNN models and lightweight CNN models. Moreover, this method does not require the modification of the internal structures of the pretrained CNN models and has high training efficiency. Experimental results on three public image classification datasets demonstrate that the proposed method outperforms the baseline traditional TL-CNN method. Compared with the baseline approach, the MDCEL approach improves the accuracy and F1 values by nearly 1–4%. In addition, an experiment on a real case dataset obtained from a manufacturing enterprise further proves the practicability of the proposed method.

Context-aware dynamic neural computational models for accurate Poly(A) signal prediction

Accurately predicting Polyadenylation (Poly(A)) signals isthe key to understand the mechanism of translation regulation and mRNA metabolism. However, existing computational algorithms fail to work well for predicting Poly(A) signals due to the vanishing gradient problem when simply increasing the number of layers. In this work, we devise a spatiotemporal context-aware neural model called ACNet for Poly(A) signal prediction based on co-occurrence embedding. Specifically, genomic sequences of Poly(A) signals are first split into k-mer sequences, and k-mer embeddings are pre-trained based on the co-occurrence matrix information; Then, gated residual networks are devised to fully extract spatial information, which has an excellent ability to control the information flow and ease the problem of vanishing gradients. The gated mechanism generates channel weights by a dilated convolution and aggregates local features by identity connections which are obtained by multi-scale dilated convolutions. Experimental results indicate that our ACNet model outperforms the state-of-the-art prediction methods on various Poly(A) signal data, and an ablation study shows the effectiveness of the design strategy.

Multi-scale dilated convolution of feature Fusion Network for Crowd counting

Multi-scale dilated convolution of feature Fusion Network for Crowd counting

Attention V-Net: A Modified V-Net Architecture for Left Atrial Segmentation

We propose a fully convolutional neural network based on the attention mechanism for 3D medical image segmentation tasks. It can adaptively learn to highlight the salient features of images that are useful for image segmentation tasks. Some prior methods enhance accuracy using multi-scale feature fusion or dilated convolution, which is basically artificial and lacks the flexibility of the model itself. Therefore, some works proposed the 2D attention gate module, but these works process 2D medical slice images, ignoring the correlation between 3D image sequences. In contrast, the 3D attention gate can comprehensively use the information of three dimensions of medical images. In this paper, we propose the Attention V-Net architecture, which uses the 3D attention gate module, and applied it to the left atrium segmentation framework based on semi-supervised learning. The proposed method is evaluated on the dataset of the 2018 left atrial challenge. The experimental results show that the Attention V-Net obtains improved performance under evaluation indicators, such as Dice, Jaccard, ASD (Average surface distance), and 95HD (Hausdorff distance). The result indicates that the model in this paper can effectively improve the accuracy of left atrial segmentation, therefore laying the foundation for subsequent work such as in atrial reconstruction. Meanwhile, our model is of great significance for assisting doctors in treating cardiovascular diseases.

Multiscale Local Enhancement Deep Convolutional Networks for the Automated 3D Segmentation of Gross Tumor Volumes in Nasopharyngeal Carcinoma: A Multi-Institutional Dataset Study.

PurposeAccurate segmentation of gross target volume (GTV) from computed tomography (CT) images is a prerequisite in radiotherapy for nasopharyngeal carcinoma (NPC). However, this task is very challenging due to the low contrast at the boundary of the tumor and the great variety of sizes and morphologies of tumors between different stages. Meanwhile, the data source also seriously affect the results of segmentation. In this paper, we propose a novel three-dimensional (3D) automatic segmentation algorithm that adopts cascaded multiscale local enhancement of convolutional neural networks (CNNs) and conduct experiments on multi-institutional datasets to address the above problems.Materials and MethodsIn this study, we retrospectively collected CT images of 257 NPC patients to test the performance of the proposed automatic segmentation model, and conducted experiments on two additional multi-institutional datasets. Our novel segmentation framework consists of three parts. First, the segmentation framework is based on a 3D Res-UNet backbone model that has excellent segmentation performance. Then, we adopt a multiscale dilated convolution block to enhance the receptive field and focus on the target area and boundary for segmentation improvement. Finally, a central localization cascade model for local enhancement is designed to concentrate on the GTV region for fine segmentation to improve the robustness. The Dice similarity coefficient (DSC), positive predictive value (PPV), sensitivity (SEN), average symmetric surface distance (ASSD) and 95% Hausdorff distance (HD95) are utilized as qualitative evaluation criteria to estimate the performance of our automated segmentation algorithm.ResultsThe experimental results show that compared with other state-of-the-art methods, our modified version 3D Res-UNet backbone has excellent performance and achieves the best results in terms of the quantitative metrics DSC, PPR, ASSD and HD95, which reached 74.49 ± 7.81%, 79.97 ± 13.90%, 1.49 ± 0.65 mm and 5.06 ± 3.30 mm, respectively. It should be noted that the receptive field enhancement mechanism and cascade architecture can have a great impact on the stable output of automatic segmentation results with high accuracy, which is critical for an algorithm. The final DSC, SEN, ASSD and HD95 values can be increased to 76.23 ± 6.45%, 79.14 ± 12.48%, 1.39 ± 5.44mm, 4.72 ± 3.04mm. In addition, the outcomes of multi-institution experiments demonstrate that our model is robust and generalizable and can achieve good performance through transfer learning.ConclusionsThe proposed algorithm could accurately segment NPC in CT images from multi-institutional datasets and thereby may improve and facilitate clinical applications.

Rubber Leaf Disease Recognition Based on Improved Deep Convolutional Neural Networks With a Cross-Scale Attention Mechanism.

Natural rubber is an essential raw material for industrial products and plays an important role in social development. A variety of diseases can affect the growth of rubber trees, reducing the production and quality of natural rubber. Therefore, it is of great significance to automatically identify rubber leaf disease. However, in practice, different diseases have complex morphological characteristics of spots and symptoms at different stages and scales, and there are subtle interclass differences and large intraclass variation between the symptoms of diseases. To tackle these challenges, a group multi-scale attention network (GMA-Net) was proposed for rubber leaf disease image recognition. The key idea of our method is to develop a group multi-scale dilated convolution (GMDC) module for multi-scale feature extraction as well as a cross-scale attention feature fusion (CAFF) module for multi-scale attention feature fusion. Specifically, the model uses a group convolution structure to reduce model parameters and provide multiple branches and then embeds multiple dilated convolutions to improve the model’s adaptability to the scale variability of disease spots. Furthermore, the CAFF module is further designed to drive the network to learn the attentional features of multi-scale diseases and strengthen the disease features fusion at different scales. In this article, a dataset of rubber leaf diseases was constructed, including 2,788 images of four rubber leaf diseases and healthy leaves. Experimental results show that the accuracy of the model is 98.06%, which was better than other state-of-the-art approaches. Moreover, the model parameters of GMA-Net are only 0.65 M, and the model size is only 5.62 MB. Compared with MobileNetV1, V2, and ShuffleNetV1, V2 lightweight models, the model parameters and size are reduced by more than half, but the recognition accuracy is also improved by 3.86–6.1%. In addition, to verify the robustness of this model, we have also verified it on the PlantVillage public dataset. The experimental results show that the recognition accuracy of our proposed model is 99.43% on the PlantVillage dataset, which is also better than other state-of-the-art approaches. The effectiveness of the proposed method is verified, and it can be used for plant disease recognition.

A multi‐scale learning method with dilated convolutional network for concrete surface cracks detection

Concrete surface cracks detection is an important task to ensure the safety of infrastructure. Because of the complexity of background and low contrast of concrete surface, it is difficult to detect the cracks on the concrete surface accurately. To tackle this problem, a multi-scale dilated convolutional method for concrete surface crack detection is proposed to improve the accuracy of detection. The proposed network is based on the encoder-decoder structure of U-Net. Cascade multi-scale dilated convolutions in the centre of the network is used to get larger receptive field without additional parameters. In the decoder stage, the feature fusion module is used to integrated the multi-scale and multi-level side network feature for the final prediction. A large crack dataset is collected as a training set and other three smaller datasets are used for evaluation. Extensive experiments have been conducted on these three crack datasets, which achieves optimal dataset scale (ODS) F-score over 0.84, optimal image scale (OIS) F-score over 0.85 and average precision (AP) over 0.86. This algorithm performs better than the current crack detection, edge detection and semantic segmentation methods.

Fully connected network with multi-scale dilation convolution module in evaluating atrial septal defect based on MRI segmentation

Background and ObjectiveAtrial septal defect (ASD) is a common congenital heart disease. During embryonic development, abnormal atrial septal development leads to pores between the left and right atria. ASD accounts for the largest proportion of congenital heart disease. Therefore, the design and implementation of an ASD intelligent auxiliary segmentation system based on deep learning segmentation of the atria has very important practical significance, which we aim to achieve in this paper. MethodsThis study proposes a multi-scale dilated convolution module, which is composed of three parallel dilated convolutions with different expansion coefficients. The original FCN network usually adopts bilinear interpolation or deconvolution methods when upsampling, both of which lead to information loss to a certain extent. In order to make up for the loss of information, it is expected that the final segmentation result can be directly connected to the deep features in the cardiac MRI. This study uses a dense upsampling convolution module, and in order to obtain the shallow position information, the original FCN jump connection module is still retained. In this research, a deep convolutional neural network for multi-scale feature extraction is designed through the multi-scale expansion convolution module. At the same time, this paper also implements two traditional machine learning segmentation methods (K-means and Watershed algorithms) and a deep learning algorithm (U-net) for comparison. ResultsThe intelligent auxiliary segmentation algorithm for atrial images proposed in this framework based on multi-scale expansion convolution and adversarial learning can achieve superior results. Among them, the segmentation algorithm based on multi-scale expansion convolution can extract the associated features of pixels in multiple ranges, and can obtain deeper feature information when using a limited downsampling layer. According to the experimental results of the multi-scale expanded convolutional network on the data set, the Proportion of Greater Contour (PGC) index of the multi-scale expanded convolutional network is 98.78, the value of Average Perpendicular Distance (ADP) is 1.72mm, and the value of Overlapping Dice Metric (ODM) is 0.935, which are higher than other models. ConclusionThe experimental results show that compared with other segmentation models, the model based on multi-scale expansion convolution has significantly improved the accuracy of segmentation. Our technique will be able to assist in the segmentation of ASD, evaluation of the extent of the defect and enhance surgical planning via atrial septal occlusion.

A Feature-Enriched Deep Convolutional Neural Network for JPEG Image Compression Artifacts Reduction and its Applications.

The amount of multimedia data, such as images and videos, has been increasing rapidly with the development of various imaging devices and the Internet, bringing more stress and challenges to information storage and transmission. The redundancy in images can be reduced to decrease data size via lossy compression, such as the most widely used standard Joint Photographic Experts Group (JPEG). However, the decompressed images generally suffer from various artifacts (e.g., blocking, banding, ringing, and blurring) due to the loss of information, especially at high compression ratios. This article presents a feature-enriched deep convolutional neural network for compression artifacts reduction (FeCarNet, for short). Taking the dense network as the backbone, FeCarNet enriches features to gain valuable information via introducing multi-scale dilated convolutions, along with the efficient 1 ×1 convolution for lowering both parameter complexity and computation cost. Meanwhile, to make full use of different levels of features in FeCarNet, a fusion block that consists of attention-based channel recalibration and dimension reduction is developed for local and global feature fusion. Furthermore, short and long residual connections both in the feature and pixel domains are combined to build a multi-level residual structure, thereby benefiting the network training and performance. In addition, aiming at reducing computation complexity further, pixel-shuffle-based image downsampling and upsampling layers are, respectively, arranged at the head and tail of the FeCarNet, which also enlarges the receptive field of the whole network. Experimental results show the superiority of FeCarNet over state-of-the-art compression artifacts reduction approaches in terms of both restoration capacity and model complexity. The applications of FeCarNet on several computer vision tasks, including image deblurring, edge detection, image segmentation, and object detection, demonstrate the effectiveness of FeCarNet further.

Fused Adaptive Receptive Field Mechanism and Dynamic Multiscale Dilated Convolution for Side-Scan Sonar Image Segmentation

Side-scan sonar (SSS) is a vital sensor for marine survey, which is widely used in military and civilian fields. The accurate segmentation of SSS images is critical in sonar image intelligent interpretation. Existing SSS image segmentation methods have several limitations, such as insufficient feature extraction, relatively worse segmentation results for tiny target categories, and serious interference by seabed reverberation noise and bright shadow region. To overcome these issues, we propose a novel encoder-decoder architecture SSS image segmentation method based on convolution neural network (CNN). First, we extract the multi-scale feature information contained in target region using the dynamic multi-scale dilated convolution (DMDC_Conv). Second, to further obtain the global and detail feature information, we construct the adaptive receptive field mechanism block (ARFM_Block). Third, we design a feature fusion attention mechanism block (FFAM_Block) to fuse high-level and low-level feature information with different scales and suppress background information interference. Final, we construct a tree structure optimization module (TSOM) to solve the problem of pixel misclassification and obtain refine SSS image segmentation results. Extensive experiments are carried out on the constructed real scene SSS image dataset. The experimental results show that the proposed method achieves 93.24% and 90.82% of MPA and MIoU, respectively, which outperforms other state-of-the-art methods and has a substantial advantage in inference speed and calculation parameters.

An Attention-Enhanced End-to-End Discriminative Network With Multiscale Feature Learning for Remote Sensing Image Retrieval

The discriminative ability of image features plays a decisive role in content-based remote sensing image retrieval (CBRSIR). However, the widely-used convolutional neural networks cannot focus on the discriminative features of important scenes, resulting in unsatisfactory retrieval performance in complex contexts. In this paper, an attention-enhanced end-to-end discriminative network with multiscale learning for CBRSIR is proposed to solve this issue. First, a multiscale dilated convolution module is embedded into some of ResNet50's residual blocks to increase the perceptual field and capture the multiscale features of remote sensing image scenes. Then, a lightweight and efficient triplet attention module is added behind each residual block to capture the salient features of remote sensing images and establish the inter-dimensional dependencies using residual transform. In addition, the end-to-end training approach is performed using an online label smoothing loss to reduce the intra-class variance of features and enhance inter-class differentiability. Experimental results on four publicly available remote sensing image datasets show that our network achieves state-of-the-art or competitive performance, especially on complex scene dataset UCMD with an average retrieval precision improvement of 3.23% to 29.35% compared to other new methods.

A novel ResNet101 model based on dense dilated convolution for image classification

Image classification plays an important role in computer vision. The existing convolutional neural network methods have some problems during image classification process, such as low accuracy of tumor classification and poor ability of feature expression and feature extraction. Therefore, we propose a novel ResNet101 model based on dense dilated convolution for medical liver tumors classification. The multi-scale feature extraction module is used to extract multi-scale features of images, and the receptive field of the network is increased. The depth feature extraction module is used to reduce background noise information and focus on effective features of the focal region. To obtain broader and deeper semantic information, a dense dilated convolution module is deployed in the network. This module combines the advantages of Inception, residual structure, and multi-scale dilated convolution to obtain a deeper level of feature information without causing gradient explosion and gradient disappearance. To solve the common feature loss problems in the classification network, the up- down-sampling module in the network is improved, and multiple convolution kernels with different scales are cascaded to widen the network, which can effectively avoid feature loss. Finally, experiments are carried out on the proposed method. Compared with the existing mainstream classification networks, the proposed method can improve the classification performance, and finally achieve accurate classification of liver tumors. The effectiveness of the proposed method is further verified by ablation experiments.HighlightsThe multi-scale feature extraction module is introduced to extract multi-scale features of images, it can extract deep context information of the lesion region and surrounding tissues to enhance the feature extraction ability of the network.The depth feature extraction module is used to focus on the local features of the lesion region from both channel and space, weaken the influence of irrelevant information, and strengthen the recognition ability of the lesion region.The feature extraction module is enhanced by the parallel structure of dense dilated convolution, and the deeper feature information is obtained without losing the image feature information to improve the classification accuracy.