Automatic crack segmentation network based on large kernel pooling transformer

An automatic blood vessel segmentation is considered as important due to the presence of similar properties between blood vessels and the early stage pathologies (i.e. Microaneurysm and hemorrhages) in retinal diseases [1,2]. Blood vessel segmentation is considered as a challenging task, if a retinal images having inadequate contrast, lightning variations, and different pathologies like exudates. In literature, there is a number of blood vessel segmentation algorithms exist, but they are not compared on similar datasets and thus, it is difficult to choose which one is best. In this paper, we have implemented five different blood vessel segmentation techniques based on Gaussian matched filter [4], edge based techniques [5,8] (Sobel, Prewitt and Kirsch) and entropy based thresholding [6] technique and compare them on similar datasets with low quality and pathological retinal images. Secondly, in low quality images, most of the pixels have lost their color intensity information and considered as noise. In order to retrieve their color information of pixels, initial pre-processing is needed. Thus, a different pre-processing approach includes mask generation and contrast enhancement to retrieve the pixel information are proposed in retinal images and compared. The results obtained are verified broadly on two publicly available standard datasets DIARETDB0 and STARE dataset comprises of 211 retinal images.

Computer-aided diagnosis is of great significance to improve the diagnostic accuracy of pancreatic cancer that has an insidious course and does not have obvious symptoms at first. However, segmentation of pancreatic cancer is challenging because the tumors vary in size with the smallest tumor having a size of about 0.5 in diameter, and most of them have irregular shapes and unclear boundaries. In this study, we developed a deep learning architecture Multi-Scale Channel Attention Unet (MSCA-Unet) for pancreatic tumor segmentation and collected CT images of 419 patients from The Affiliated Hospital of Qingdao University and a public dataset. We embedded the multi-scale network into the encoder to extract semantic information at different scales and the decoder to provide supplemental information to overcome the loss of information in the upsampling and the drift of the localized tumor due to the upsampling and skip connections. We adopted the channel attention unit after the multi-scale convolution to emphasize the informative channels, which was observed to have the effects of accelerating the positioning process, reducing false positives, and improving the accuracy of outlining very small, irregular pancreatic tumors. Our results show that our network outperformed the other current mainstream segmentation networks and achieved a Dice index of 68.03%, a Jaccard of 59.31%, and an FPR of 1.36% on the private dataset Task-01 without data pre-processing. Compared with the other pancreatic tumor segmentation networks on the public dataset Task-02, our network produced the best Dice index, 80.12%, with the assistance of the data pre-processing scheme. This study strategically utilizes the multi-scale convolution and channel attention mechanism of the architecture to provide a dedicated network for segmentation of small and irregular pancreatictumors.

Based on the theory and application, this paper discusses the optimization of art image segmentation algorithm based on FFNN (Feed Forward Neural Network). In this paper, residual units are used in the corresponding stages of encoder and decoder, and feature information of several convolution layers in each convolution stage of encoder is extracted at the same time. And the feature pyramid module is used to extract multiscale features from the feature map of the last convolution stage in the encoder. Finally, pixel by pixel additions combine the previously mentioned feature information into the corresponding layer of the decoder. Additionally, an improved weight adaptive algorithm based on feature preservation is suggested in this paper, which addresses the issue that the conventional image segmentation algorithm is noise-sensitive. The adaptive connection weight mechanism is also introduced. The accuracy and recall rates of this optimization algorithm can both reach 96.574%, according to the results of 50% cross-validation. All the segmentation performance evaluation indexes of this algorithm are higher than the existing main algorithms. Moreover, the algorithm takes a short time, does not need too much manual intervention, and can effectively segment artistic images. The optimization algorithm in this paper has certain reference significance for the related research of artistic image segmentation.

To apply deep convolution neural network to the segmentation task in myocardial arterial spin labeled perfusion imaging and to develop methods that measure uncertainty and that adapt the convolution neural network model to a specific false-positive versus false-negative tradeoff. The Monte Carlo dropout U-Net was trained on data from 22 subjects and tested on data from 6 heart transplant recipients. Manual segmentation and regional myocardial blood flow were available for comparison. We consider 2 global uncertainty measures, named "Dice uncertainty" and "Monte Carlo dropout uncertainty," which were calculated with and without the use of manual segmentation, respectively. Tversky loss function with a hyperparameter β was used to adapt the model to a specific false-positive versus false-negative tradeoff. The Monte Carlo dropout U-Net achieved a Dice coefficient of 0.91 ± 0.04 on the test set. Myocardial blood flow measured using automatic segmentations was highly correlated to that measured using the manual segmentation (R2 = 0.96). Dice uncertainty and Monte Carlo dropout uncertainty were in good agreement (R2 = 0.64). As β increased, the false-positive rate systematically decreased and false-negative rate systematically increased. We demonstrate the feasibility of deep convolution neural network for automatic segmentation of myocardial arterial spin labeling, with good accuracy. We also introduce 2 simple methods for assessing model uncertainty. Finally, we demonstrate the ability to adapt the convolution neural network model to a specific false-positive versus false-negative tradeoff. These findings are directly relevant to automatic segmentation in quantitative cardiac MRI and are broadly applicable to automatic segmentation problems in diagnostic imaging.

Automatic three-dimensional analysis of posterosuperior full-thickness rotator cuff tear size on MRI

In the task of PolSAR image classification, effectively utilizing convolutional neural networks and vision transformer models with limited labeled data poses a critical challenge. This article proposes a novel method for PolSAR image classification that combines channel-wise convolutional local perception, detachable self-attention, and a residual feedforward network. Specifically, the proposed method comprises several key modules. In the channel-wise convolutional local perception module, channel-wise convolution operations enable accurate extraction of local features from different channels of PolSAR images. The local residual connections further enhance these extracted features, providing more discriminative information for subsequent processing. Additionally, the detachable self-attention mechanism plays a pivotal role: it facilitates effective interaction between local and global information, enabling the model to comprehensively perceive features across different scales, thereby improving classification accuracy and robustness. Subsequently, replacing the conventional feedforward network with a residual feedforward network that incorporates residual structures aids the model in better representing local features, further enhances the capability of cross-layer gradient propagation, and effectively alleviates the problem of vanishing gradients during the training of deep networks. In the final classification stage, two fully connected layers with dropout prevent overfitting, while softmax generates predictions. The proposed method was validated on the AIRSAR Flevoland, RADARSAT-2 San Francisco, and RADARSAT-2 Xi’an datasets. The experimental results demonstrate that the proposed method can attain a high level of classification performance even with a limited amount of labeled data, and the model is relatively stable. Furthermore, the proposed method has lower computational costs than comparative methods.

Skin cancer is the most common form of cancer worldwide, with early detection being critical to reducing mortality and minimizing the need for invasive therapies. This thesis proposes an automatic and low-cost skin cancer detection system using active infrared thermography (IRT), machine learning and deep learning. Active IRT leverages controlled thermal stimuli to identify malignant lesions based on unique thermal regulation curves (TRCs) observed in tumor-affected areas. The study introduces a novel detection scheme that integrates spatio-temporal features extracted from TRCs, visible imagery, and machine learning classifiers. Along with this, the use of a U-Net convolutional neural network for automatic lesion segmentation, trained on high-quality datasets and tested in real clinical conditions, is proposed. Additionally, the research evaluates a range of low-cost infrared cameras and develops a video degradation model to simulate their per formance. This evaluation showed that the proposed detection algorithm using expert segmentation allows working with low-cost cameras with an accuracy of approximately 85% with mid-range infrared cameras and 83% with a low-end camera. Finally, the application of transfer learning to TRCs was explored, with a proposed strategy to generate images from the curves. Using a ResNet152V2 network, the curves were classified as benign or malignant, and the average ma lignancy of the analyzed TRC set was calculated. Incorporating this new feature, the detection system achieved an accuracy of 90.86% with expert segmentation and 89.20% with automatic segmentation. These results suggest that incorpo rating this feature into the detection algorithm for lower-quality cameras could enhance its performance, potentially matching the accuracy achieved with high quality video processing.

Attention-guided multi-scale learning network for automatic prostate and tumor segmentation on MRI

Cardiac Magnetic Resonance (CMR) imaging is widely used in the clinic to assess the patient-specific cardiac structure and function. However, the manual analysis of the CMR data is tedious and subjective. In this work, we developed a fully automatic segmentation system for the left ventricle (LV) myocardium from MR cine images. The system consists of three major components. Firstly, a conventional convolutional neural network (CNN) was trained to detect the global region of interest (ROI) of LV. Secondly, a novel multi-scope CNN was proposed to segment the LV myocardium from the reduced ROI, taking advantage of the image context in different scopes, such that both the local accuracy and global consistency can be implicitly learned by the CNN. Finally the results were pruned with simple morphological filtering preserving the largest component. With a relatively small training set of 200 MR cine images, the method achieved an average segmentation accuracy of 0.71 as expressed by the Dice overlap index. The proposed method can be applied to automatically segment the LV from MR images with the reasonable accuracy, or as a proper initialization for local shape methods to achieve further refined results.

Retinal vessel segmentation is an essential part of ocular disease diagnosis. However, due to complex vascular structure, large‐scale variations of retinal vessels, as well as inefficiency of vessel segmentation speed, accurate and fast automatic vessel segmentation for retinal images is still technically challenging. To tackle these issues, we present a lightweight multi‐scale‐aware network (LMSA‐Net) for retinal vessel segmentation. The network leverages the encoder‐decoder structure that was used in U‐Net. In the encoder, we propose a ghosted sandglass residual (GSR) block, aiming at greatly reducing the parameters and computational cost while obtaining richer semantic information. After that, a multi‐scale feature‐aware aggregation (MFA) module is designed to perceive multi‐scale semantic information for effective information extraction. Then, a global adaptive upsampling (GAU) module is proposed to guide the effective fusion of high‐ and low‐level semantic information in the decoder. Experiments are conducted on three public datasets, including DRIVE, CHASE_DB1, and STARE. The experimental results indicate the effectiveness of the LMSA‐Net, which can achieve better segmentation performance than other state‐of‐the‐art methods.

Segmentation of stroke lesions plays an important role in the diagnosis and treatment of computer-aided brain diseases. However, due to the complex relationships between the modes, convolutional neural networks (CNN) using a simple multi-modal fusion strategy are difficult to learn complex and nonlinear complementary information between the modes. On the other hand, with the rapid development of CNN, codec structure has shown great potential in the field of medical image segmentation, but it also exposes corresponding problems. Two-dimensional CNN ignores the information between the slice sequences of medical images, while three-dimensional CNN has a high demand for computing resources. At the same time, the undifferentiated feature aggregation can easily lead to the problem of feature dislocation. To solve these problems, this paper proposes an automatic segmentation network based on multi-mode multi-slice cooperation. In this network, we first design a multi-branch parallel processing strategy, and divide the network into three parallel parts according to different processing data, so as to realize collaborative learning among multi-slices innovatively. Secondly, a convolutional LSTM module with attention mechanism is introduced to directly use the correlation between slices. Third, a cross-modal feature interaction module is developed to adaptively and efficiently fuse and refine multi-modal features. The proposed algorithm was validated on the public stroke dataset ISLES2015, and the experimental results show that the proposed algorithm achieves accurate and effective multimodal stroke segmentation.

Multispectral remote sensing images are often contaminated by haze, which causes low image quality. In this paper, a novel dehazing method based on a deep convolutional neural network (CNN) with the residual structure is proposed for multispectral remote sensing images. First, multiple CNN individuals with the residual structure are connected in parallel and each individual is used to learn a regression from the hazy image to the clear image. Then, the outputs of CNN individuals are fused with weight maps to produce the final dehazing result. In the designed network, the CNN individuals, mining multiscale haze features through multiscale convolutions, are trained using different levels of haze samples to achieve different dehazing abilities. In addition, the weight maps change with the haze distribution, and the fusion of the CNN individuals is adaptive. The designed network is end-to-end, and putting a hazy image into it, the clear scene can be restored. To train the network, a wavelength-dependent haze simulation method is proposed to generate labeled data, which can synthesize hazy multispectral images highly close to real conditions. Experimental results show that the proposed method can accurately remove the haze in each band of multispectral images under different scenes.

Most of the semantic segmentation real‐time networks improve the segmentation speed by reducing the spatial resolution, leading to the accuracy being significantly reduced as a result. To solve this problem, we propose feature enhancement module (FEM), feature extraction and fusion module (FEFM). By extracting and enhancing the future map before the image down‐sample on the backbone and fusing the low‐level features with rich details and the high‐level features with more semantic information. Based on the FEM and FEFM, we introduce a real‐time semantic segmentation network feature extraction and enhancement network. In the experiment, using Cityscapes and CamVid datasets, the proposed network achieves a balance between computing speed and accuracy. Without additional processing and pretraining, it achieves 75.47% Mean IoU on the Cityscapes test dataset with only 29.96G Flops and a speed of 94 frames per second on a single RTX 2080Ti card. Code is available at https://github.com/favoMJ/FEENet.

In the harsh underground environment, the transmission gear in the rocker arm of the shearer is susceptible to impact load, frequently leading to failures that result in economic losses or casualties. Therefore, real-time and accurate gear fault diagnosis is crucial. Existing diagnostic models struggle with insufficient feature extraction and low diagnostic accuracy when addressing nonlinear, nonsmooth signals with strong noise in this environment. Therefore, an innovative parallel integration model of Swin Transformer and multiscale convolution is proposed to extract both global and local features under different fault modes and noise conditions. A spatial interaction block (SIB) is introduced into the Swin Transformer branch to enhance global information acquisition. In multiscale convolutional neural network (CNN) branches, depthwise separable convolutions with different expansion coefficients are utilized for multiscale extraction of local features to reduce computational complexity. Considering the differences between global and local features, a feature interaction module (FIM) is added between the two branches to realize the interaction between local and global features, considerably improving the model diagnosis performance. Furthermore, to meet the high real-time diagnosis requirements, a cloud-edge collaboration (CEC) framework is designed to realize the timely alarm of gear faults through the efficient collaboration between edge nodes and the industrial cloud platform. Using the vibration data collected from the rocker arm loading test bed of Taizhong Coal Machinery and the gear data provided by Southeast University for validation, the results show that the diagnostic precision of the proposed method can reach up to 99.55%. Even when the signal-to-noise ratio (SNR) is −2 db, the accuracy is still 93.67%, which is superior to other advanced comparison models, and significantly reduces the completion time of fault diagnosis tasks, and has strong anti-noise and real-time diagnosis capabilities.

Skin wound segmentation in photographs allows non-invasive analysis of wounds that supports dermatological diagnosis and treatment. In this paper, we propose a novel feature augment network (FANet) to achieve automatic segmentation of skin wounds, and design an interactive feature augment network (IFANet) to provide interactive adjustment on the automatic segmentation results. The FANet contains the edge feature augment (EFA) module and the spatial relationship feature augment (SFA) module, which can make full use of the notable edge information and the spatial relationship information be-tween the wound and the skin. The IFANet, with FANet as the backbone, takes the user interactions and the initial result as inputs, and outputs the refined segmentation result. The pro-posed networks were tested on a dataset composed of miscellaneous skin wound images, and a public foot ulcer segmentation challenge dataset. The results indicate that the FANet gives good segmentation results while the IFANet can effectively improve them based on simple marking. Comprehensive comparative experiments show that our proposed networks outperform some other existing automatic or interactive segmentation methods, respectively.