Pathological images hold significant importance for disease diagnosis. When employing artificial intelligence deep learning for pathological image segmentation, numerous challenges arise. For example, difficulties in data collection and annotation, substantial individual variations affecting training, and the lack of interpretability in model decisions, which leads to low levels of physician trust. This study proposes the MSDAUnet segmentation method. Firstly, an image denoising and enhancement module composed of a Dynamic Residual Attention Network (DRAN) and an image enhancement method based on Improved Color Histogram Equalization (MCHE) is constructed to remove image noise and improve image clarity. Then, a Dual Attention Module (DAM) is proposed, which connects the Squeeze-andExcitation Block (SE Block) and the Spatial Attention Module (SAM) in series to extract information from both the channel and spatial dimensions and obtain key features. Finally, the Atrous Spatial Pyramid Pooling (ASPP) module is introduced to obtain multi-scale information.The combination of these modules is used for automatic pathological image segmentation. Compared with the traditional and typical UNet model, the segmentation effect of MSDAUNet is more excellent. On the Monash University dataset, its IoU index reaches$\mathbf{7 2. 7 \%}$, nearly$\mathbf{7 \%}$higher than UNet, and the DSC index is 84.9 %, also about 7 % higher than UNet. It vigorously promotes the development of medical imaging processing technology and provides more accurate AI assistance for medical decision-making.

Artificial intelligence has become an indispensable element in modern clinical diagnostics, offering strong potential for assisting in tumor assessment. Gliomas, among the most malignant brain tumors, are difficult to delineate precisely because of their irregular geometry and indistinct tumor-tissue boundaries. To enhance diagnostic reliability, this study presents a Glioma Diagnostic Decision-Making System (GDDMS) incorporating a Feature Fusion Bridge (FFB) within a Transformer-based segmentation framework. The FFB enables multi-scale integration of encoder features, effectively coupling global context with localized structural details to refine tumor boundary detection. Comprehensive experiments demonstrate that GDDMS achieves a Dice coefficient of 0.892, outperforming other comparative methods. These findings suggest that the system can provide dependable segmentation support and improve diagnostic efficiency in clinical glioma evaluation.

AI medical imaging diagnosis enables physicians to rapidly acquire and analyze image information using deep learning techniques. Consequently, it holds a significant position in medical decision-making assistance. Specifically, the complexity and large scale of digital pathology images necessitate accurate recognition of cell arrangement, size, and shape, which aids physicians in identifying cancerous regions. However, existing medical image segmentation techniques often fail to deliver sufficiently detailed results in unknown target domains, resulting in the blurring of cell nucleus boundaries. In order to introduce more display-like information to solve the problem of obscure cell nucleus boundaries, this study proposes an accurate recognition method (UNW) for cell nucleus recognition in medical decision-making systems. It solves the problem of the distribution difference between the source domain and the unknown target domain by inserting feature center alignment (FCA) and feature distribution alignment (FDA) modules into the encoder and decoder. UNW performs instance normalization (IN) and instance whitening (IW) on the feature images to better identify the feature detail information, which results in more distinct boundaries between the cell nucleus and other tissues. Experimental results on various backbone networks show that our method can obtain more accurate decision results. Among them, the IOU reaches 0.704, which outperforms existing models, while the FCA and FDA maintain the segmentation ability of the model in the face of test data that is inconsistent with the distribution of the training data.

Abstract Artificial intelligence decision systems play an important supporting role in the field of medical information. Medical image analysis is an important part of decision systems and an even more important part of medical diagnosis and treatment. The wealth of cellular information in histopathological images makes them a reliable means of diagnosing tumors. However, due to the large size, high resolution, and complex background structure of pathology images, deep learning methods still have various difficulties in the recognition of pathology images. Based on this, this study proposes a two‐stage continuous improvement‐based approach for pathology image recognition in medical decision systems. For pathology images with complex backgrounds, normalization and enhancement is performed to remove the effects of noise color, and light‐dark inconsistencies on the segmentation network. The continuous refinement PSP Net (CRPSPNet) is then designed for accurate recognition of the pathology images. CRPSPNet is divided into two stages: Pyramid Scene Parsing Network segmentation to obtain coarse segmentation results; and continuous refinement model refines the results of the first stage. Experiments using more than 1,000 osteosarcoma pathology images have shown that the method gives more accurate results with fewer computer resources and processing time than traditional optimization models. Its Intersection over Union achieves 0.76.

Medical images play a significant part in biomedical diagnosis, but they have a significant feature. The medical images, influenced by factors such as imaging equipment limitations, local volume effect, and others, inevitably exhibit issues like noise, blurred edges, and inconsistent signal strength. These imperfections pose significant challenges and create obstacles for doctors during their diagnostic processes. To address these issues, we present a pathology image segmentation technique based on the multiscale dual attention mechanism (MSDAUnet), which consists of three primary components. Firstly, an image denoising and enhancement module is constructed by using dynamic residual attention and color histogram to remove image noise and improve image clarity. Then, we propose a dual attention module (DAM), which extracts messages from both channel and spatial dimensions, obtains key features, and makes the edge of the lesion area clearer. Finally, capturing multiscale information in the process of image segmentation addresses the issue of uneven signal strength to a certain extent. Each module is combined for automatic pathological image segmentation. Compared with the traditional and typical U‐Net model, MSDAUnet has a better segmentation performance. On the dataset provided by the Research Center for Artificial Intelligence of Monash University, the IOU index is as high as 72.7%, which is nearly 7% higher than that of U‐Net, and the DSC index is 84.9%, which is also about 7% higher than that of U‐Net.

Telematics technology development offers vehicles a range of intelligent and convenient functions, including navigation and mapping services, intelligent driving assistance, and intelligent traffic management. However, since these functions deal with sensitive information like vehicle location and driving habits, it is crucial to address concerns regarding information security and privacy protection. Federated learning (FL) is highly suitable for addressing such problems due to its characteristics, in which a client does not need to share private data and upload model parameters to a parameter server via the network. This results in the establishment of a federated vehicle network (FVN). As a distributed paradigm, the efficiency of communication is crucial in federated learning as it impacts all aspects of the FVN. This paper introduces a parameter freezing algorithm based on historical information to reduce the data transferred between the client and the parameter server in each round of communication, thus minimizing the communication overhead of federated learning. Additionally, we propose using a particle swarm algorithm to allocate network bandwidth to each vehicle based on the packet sizes sent by each vehicle (i.e., the non-freezing parameters) to minimize the communication latency in each FL round. Furthermore, due to the high time complexity of the particle swarm algorithm, we employ it to generate training data for training a transformer model with fast response and sufficient accuracy, thereby accelerating the bandwidth allocation process. Through extensive experiments, we prove the feasibility of our approach and its efficiency in improving communication in federated learning.

Digital pathology images' extensive cellular information provide a trustworthy foundation for tumor diagnosis. With the aid of computer-aided diagnostics, pathologists can locate crucial information more quickly. The cascade structure refines the segmentation results by utilizing its multi-task and multi-stage characteristics. However, cascade-based models require downsampling and cropping of patches during the inference process due to the ultra-high resolution and complex structure of pathology images. This not only increases the cost and computation time but also results in the loss of cellular details and corrupts the global contextual information. This study proposes a Digital Pathology Image Assistance Program (CRSDPI) for medical decision-making systems that is based on continuous improvement. After locating the region of interest using the maximum inter-class variance method, the pictures are preprocessed to account for the impacts of staining inconsistencies and sensitivity variations on the model's performance. Ultimately, we create a two-phase continuously refined segmentation network (TCRNet) by combining an enhanced continuous refinement model with a coarse segmentation network built on a pyramid scene parsing network. The coarse segmentation network introduces an auxiliary loss term to speed up convergence, and the refined model introduces an implicit function to reduce computational cost and reconstruct more details. The TCRNet model refines the target by successively aligning the features without the need to take cascading decoder operations after encoder. Experiments conducted on digital pathology images of breast cancer and osteosarcoma demonstrate the superior prediction accuracy and computational speed of our strategy.

Abstract The network transmission speed has been greatly improved, thanks to the power of 5G technology. The millisecond‐level communication delay has made a qualitative leap in communication quality. However, the sharp increase in the number of nodes connected to the internet has resulted in an explosion of traffic. Ensuring stable network transmission in the face of large data volumes has become an urgent problem to be solved. Existing research mainly optimizes for low data volumes of nodes, and cannot dynamically adapt to node transmission load to cope with explosive growth. This is because the node's cache is always at a high level, which causes community communication to be blocked. To address this issue, the authors have designed an effective data communication community establishment scheme. By adding a data information flow (DIF) attribute to each user and using the dynamic feedback adjustment (DFA) mechanism, the authors can combine the entire communication community and control the user's data information flow concentration at a relatively low level. This degree of optimization ensures that the user node maintains a relatively stable state for fast communication. The authors design an effective data communication community establishment scheme. By adding a data information flow (DIF) attribute to each user, using the dynamic feedback adjustment mechanism and combining the entire communication community, the user's data information flow concentration is controlled at a relatively low level.

Abstract Due to the improvement in computing power and the development of computer technology, deep learning has pene‐trated into various fields of the medical industry. Segmenting lesion areas in medical scans can help clinicians make accurate diagnoses. In particular, convolutional neural networks (CNNs) are a dominant tool in computer vision tasks. They can accurately locate and classify lesion areas. However, due to their inherent inductive bias, CNNs may lack an understanding of long‐term dependencies in medical images, leading to less accurate grasping of details in the images. To address this problem, we explored a Transformer‐based solution and studied its feasibility in medical imaging tasks (OstT). First, we performed super‐resolution reconstruction on the original MRI image of osteosarcoma and improved the texture features of the tissue structure to reduce the error caused by the unclear tissue structure in the image during model training. Then, we propose a Transformer‐based method for medical image segmentation. A gated axial attention model is used, which augments existing architectures by introducing an additional control mechanism in the self‐attention module to improve segmentation accuracy. Experiments on real datasets show that our method outper‐forms existing models such as Unet. It can effectively assist doctors in imaging examinations.

Osteosarcoma is the most common malignant bone tumor with a high degree of malignancy and misdiagnosis rates. Pathological images are crucial for its diagnosis. However, underdeveloped regions currently lack sufficient high-level pathologists, leading to uncertain diagnostic accuracy and efficiency. Existing research on pathological image segmentation often neglects the differences in staining styles and lack of data, without considering medical backgrounds. To alleviate the difficulty in diagnosing osteosarcoma in underdeveloped areas, an intelligent assisted diagnosis and treatment scheme for osteosarcoma pathological images, ENMViT, is proposed. ENMViT utilizes KIN to achieve normalization of mismatched images with limited GPU resources and uses traditional data enhancement methods, such as cleaning, cropping, mosaic, Laplacian sharpening, and other techniques to alleviate the issue of insufficient data. A multi-path semantic segmentation network combining Transformer and CNN is used to segment images, and the degree of edge offset in the spatial domain is introduced into the loss function. Finally, noise is filtered according to the size of the connecting domain. This article experimented on more than 2000 osteosarcoma pathological images from Central South University. The experimental results demonstrate that this scheme performs well in each stage of the osteosarcoma pathological image processing, and the segmentation results' IoU index is 9.4% higher than the comparative models, demonstrating its significant value in the medical industry.

Osteosarcoma is one of the common malignant bone tumors, which is great harm to human health. Magnetic resonance imaging (MRI) is a common method for osteosarcoma detection because it can more clearly see the extent of lesions and the degree of invasion, with minimal damage to the human body. However, in developing countries or areas with less developed medical technology, due to the high cost of hardware equipment and the problem of image detection accuracy, the existing MRI image-assisted detection technology still has not achieved the expected application. Based on this, this study proposes an attention-based MRI image-assisted segmentation system for osteosarcoma (AIMSost). The results of the system segmentation can assist doctors to determine the tumor area and identify the tumor location. Among them, we design a segmentation method, which is based on the idea of AttendSeg. We design a new module to replace the original convolutional layer, and use residual network structure, LayerNorm, and Depth-Wise to reduce the number of parameters and operational complexity of the model. Experiments show that the DSC and IOU values of AIMSost are improved by more than 2.79% and 2.03%, respectively, compared with other models in the experiment. Our proposed method has not only higher segmentation accuracy, but also higher efficiency and a more lightweight model structure. The code for AIMSost is available online at https://github.com/gnafuog/AIMSost.git.

Nowadays, the most common primary bone tumor is osteosarcoma, which mostly occurs in teenagers. A common diagnosis method is currently that doctors manually diagnose osteosarcoma in magnetic resonance imaging (MRI) images because it is nonradioactive and has no biological damage to brain tissue and more obvious performance in soft tissue components such as tumors, blood vessels, and muscles in MRI images. However, this method is labor-intensive and time-consuming work, and cannot guarantee the accuracy of the diagnostic results. Existing osteosarcoma MRI image segmentation methods either aim to model the global context to improve the inner consistency of objects, or multiscale feature fusion to refine the detail of objects along their boundaries, which all ignore the interaction between the body of the object and the object boundary. Therefore, this paper proposes a novel segmentation method for osteosarcoma MRI images based on DecoupleSegNet, which explores the relationship between body feature and edge feature. It can assist doctors in diagnosing osteosarcoma and improve their work efficiency. First, we warp the feature of MRI images through learning a flow field so we can make the object more consistent. We then make further work to optimize the resulting body feature and residual edge feature through explicitly sampling pixels from different parts under decoupled supervision. Through these steps, we finally obtain the final feature map with fine boundaries from the MRI image of osteosarcoma. We take a test by using more than 80,000 osteosarcoma MRI images obtained from three hospitals in China. We find that compared with existing osteosarcoma MRI image segmentation methods, our proposed method achieves 90.51 Intersection of Union % with few parameters on the test, outperforming other models. In the test, we prove that our proposed method has better accuracy and lower resource consumption.

Osteosarcoma is a malignant bone tumor commonly found in adolescents or children, with high incidence and poor prognosis. Magnetic resonance imaging (MRI), which is the more common diagnostic method for osteosarcoma, has a very large number of output images with sparse valid data and may not be easily observed due to brightness and contrast problems, which in turn makes manual diagnosis of osteosarcoma MRI images difficult and increases the rate of misdiagnosis. Current image segmentation models for osteosarcoma mostly focus on convolution, whose segmentation performance is limited due to the neglect of global features. In this paper, we propose an intelligent assisted diagnosis system for osteosarcoma, which can reduce the burden of doctors in diagnosing osteosarcoma from three aspects. First, we construct a classification-image enhancement module consisting of resnet18 and DeepUPE to remove redundant images and improve image clarity, which can facilitate doctors' observation. Then, we experimentally compare the performance of serial, parallel, and hybrid fusion transformer and convolution, and propose a Double U-shaped visual transformer with convolution (DUconViT) for automatic segmentation of osteosarcoma to assist doctors' diagnosis. This experiment utilizes more than 80,000 osteosarcoma MRI images from three hospitals in China. The results show that DUconViT can better segment osteosarcoma with DSC 2.6% and 1.8% higher than Unet and Unet++, respectively. Finally, we propose the pixel point quantification method to calculate the area of osteosarcoma, which provides more reference basis for doctors' diagnosis.

Osteosarcoma is the most common malignant osteosarcoma, and most developing countries face great challenges in the diagnosis due to the lack of medical resources. Magnetic resonance imaging (MRI) has always been an important tool for the detection of osteosarcoma, but it is a time-consuming and labor-intensive task for doctors to manually identify MRI images. It is highly subjective and prone to misdiagnosis. Existing computer-aided diagnosis methods of osteosarcoma MRI images focus only on accuracy, ignoring the lack of computing resources in developing countries. In addition, the large amount of redundant and noisy data generated during imaging should also be considered. To alleviate the inefficiency of osteosarcoma diagnosis faced by developing countries, this paper proposed an artificial intelligence multiprocessing scheme for pre-screening, noise reduction, and segmentation of osteosarcoma MRI images. For pre-screening, we propose the Slide Block Filter to remove useless images. Next, we introduced a fast non-local means algorithm using integral images to denoise noisy images. We then segmented the filtered and denoised MRI images using a U-shaped network (ETUNet) embedded with a transformer layer, which enhances the functionality and robustness of the traditional U-shaped architecture. Finally, we further optimized the segmented tumor boundaries using conditional random fields. This paper conducted experiments on more than 70,000 MRI images of osteosarcoma from three hospitals in China. The experimental results show that our proposed methods have good results and better performance in pre-screening, noise reduction, and segmentation.

Lung cancer has the highest mortality rate among all malignancies. Non-micro pulmonary nodules are the primary manifestation of early-stage lung cancer. If patients can be detected with nodules in the early stage and receive timely treatment, their survival rate can be improved. Due to the large number of patients and limited medical resources, doctors take a longer time to make a diagnosis, which reduces efficiency and accuracy. Besides, there are no suitable approaches for developing countries. Therefore, we propose a 2.5D-based cascaded multi-stage framework for automatic detection and segmentation (DS-CMSF) of pulmonary nodules. The first three stages of the framework are used to discover lesions, and the latter stage is used to segment them. The first locating stage introduces the classical 2D-based Yolov5 model to locate the nodules roughly on axial slices. The second aggregation stage proposes a candidate nodule selection (CNS) algorithm to locate further and reduce redundant candidate nodules. The third classification stage uses a multi-size 3D-based fusion model to accommodate nodules of varying sizes and shapes for false-positive reducing. The last segmentation stage introduces multi-scale and attention modules into 3D-based UNet autoencoder to segment the nodular regions finely. Our proposed framework achieves 95.95% sensitivity and 89.50% CPM for nodules detection on the LUNA16 dataset, and 86.75% DSC for nodules segmentation on the LIDC-IDRI dataset. Moreover, our approach also achieves the accuracy-complexity trade-off, which can effectively realize the auxiliary diagnosis of pulmonary nodules in developing countries.