This study proposed a novel extraction method of c-Fos protein regions in DAB(3,3‘-diaminobenzidine)-stained mouse brain slice images using the U-Net model combined with the multi-channelization and $1\times 1$ con...
详细信息
ISBN:
(数字)9798350373332
ISBN:
(纸本)9798350373349
This study proposed a novel extraction method of c-Fos protein regions in DAB(3,3‘-diaminobenzidine)-stained mouse brain slice images using the U-Net model combined with the multi-channelization and
$1\times 1$
convolution techniques. Our U-Net m odel, called
$1\times 1$
conv U-Net, was applied to three DAB- strainedmouse slice images whose resolutions were
$1000\mathrm{x}1000$
pixels. The experimental results demonstrated that the
$1\times 1$
conv U-Net outperformed the conventional U-Net.
This research provides a revolutionary artificial intelligence (AI) and cloud-based system for assessing the severity of sports injuries in real time and tracking healing progress. The objective is to develop an AI an...
详细信息
ISBN:
(数字)9798331521394
ISBN:
(纸本)9798331521400
This research provides a revolutionary artificial intelligence (AI) and cloud-based system for assessing the severity of sports injuries in real time and tracking healing progress. The objective is to develop an AI and Cloud-based system for evaluating sports injuries in real time, providing accurate severity assessment, recovery tracking, and personalised treatment recommendations with powerful Natural Language Processing (NLP) and image analysis algorithms. The proposed method uses modern NLP and deep learning-based image analysis algorithms to process and analyze data from a variety of sources, such as IoT-enabled wearable devices, medical reports, and athlete feedback. NLP is used to extract useful insights from unstructured text, while image analysis finds damage patterns in medical images. By combining these technologies with cloud computing, the system provides a scalable and economical solution for real-time injury assessment, including personalized rehabilitation programs and predictive modelling. AI enables automated, data-driven decision-making, which reduces human error while increasing diagnostic speed and accuracy. Future study will look at the addition of more AI algorithms, expanding the dataset to include more injury kinds, and improving the system's real-time feedback capability. This method aims to transform sports injury care and recovery optimization.
Generalized zero-shot learning (ML-GZSL) has demonstrated significant potential in medical diagnostics due to doctors' need to process large volumes of medical images. Vision Transformers (ViTs), due to their Tran...
详细信息
ISBN:
(数字)9798331513054
ISBN:
(纸本)9798331513061
Generalized zero-shot learning (ML-GZSL) has demonstrated significant potential in medical diagnostics due to doctors' need to process large volumes of medical images. Vision Transformers (ViTs), due to their Transformer-like structure, are considered to have superior feature-generation capabilities in cross-text-image tasks. BioMedBERT, based on the BERT architecture and domain-specific pre-training for biomedical natural language processing, is considered to have significant label embedding capabilities in cross-text-image tasks. In this paper, we propose MMKNet, a novel method that employs ViTs to construct global and local features of images for visual knowledge from images while using BioMedBERT with prompt tuning for the label embedding to achieve the knowledge from textual embedding in biomedical corpora. To integrate multi-modal information, we design a unique combined decision layer, which outputs similarity scores between images and class labels, providing the predicted classifications. Our method is class-independent during inference, enabling the model to predict unseen classes. Experiments on the NIH-ChestXray14, Kaggle retina, and Multi-Label Retinal Diseases (MuReD) datasets demonstrate that our method outperforms baseline models across multiple performance metrics, which can potentially optimize doctors' workflows by allowing them to focus on diagnosing complex cases, addressing challenges of limited dataset sizes and incomplete disease coverage in the medical imaging domain.
With multiple hidden layers and massive combinations of features and weights, deep learning models are hard to understand, and even more difficult to interact with. In this paper we describe a visual analytics platfor...
详细信息
This article presents the development of a cross-platform mobile application using the Google’s Gemini Generative Artificial Intelligence in Flutter, aimed at providing support to people with visual impairments in un...
详细信息
ISBN:
(数字)9798350363043
ISBN:
(纸本)9798350363050
This article presents the development of a cross-platform mobile application using the Google’s Gemini Generative Artificial Intelligence in Flutter, aimed at providing support to people with visual impairments in understanding visual content. The application leverages mobile computing technologies to offer an accessible and intuitive experience, allowing users to obtain real-time image descriptions or from the gallery. Addition- ally, it integrates the Google Gemini artificial intelligence model, which utilizes computer vision techniques to generate contextual descriptions of images. The article describes the architecture of the application, the main components of its implementation, and configuration of the model, as well as the challenges faced during the development process. Ultimately, it seeks to understand the benefits of using Gemini and ways to configure and receive the new model from Google via API.(Abstract)
Recent advancements in medical image processing have significantly enhanced the capabilities of diagnostic imaging and treatment planning. This paper provides a comprehensive overview of the latest methodologies and t...
详细信息
ISBN:
(数字)9798350378092
ISBN:
(纸本)9798350378108
Recent advancements in medical image processing have significantly enhanced the capabilities of diagnostic imaging and treatment planning. This paper provides a comprehensive overview of the latest methodologies and technologies in image processing that contribute to more precise diagnoses and improved treatment strategies. We explore several key areas including deep learning algorithms for image segmentation, enhancement techniques for improving image clarity and detail, and innovative approaches for 3D visualization. Furthermore, we discuss the integration of artificial intelligence with traditional imaging techniques to accelerate decision-making processes and personalize patient care. The impact of these advancements is evaluated through case studies and clinical feedback, which demonstrate improved outcomes in patient diagnosis and treatment planning. Our findings suggest that the integration of advanced image processing techniques can transform medical imaging into a more dynamic tool for healthcare professionals, ultimately leading to better patient outcomes.
Traditional neuro-rehabilitation methods often do not effectively involve participants, mainly because they lack a truly immersive experience. The present study aims to address this issue by combining brain-computer i...
详细信息
ISBN:
(数字)9798350385922
ISBN:
(纸本)9798350385939
Traditional neuro-rehabilitation methods often do not effectively involve participants, mainly because they lack a truly immersive experience. The present study aims to address this issue by combining brain-computer interfaces (BCI) and virtual reality (VR) technologies with customized 3D avatars. This approach takes advantage of the brain’s neural processes, stimulating areas responsible for carrying out movements when visualizing them. The heart of this system lies in combining the immersive capabilities of VR with the precision of electroencephalography to record brain activity during motor imagery (MI). Customized 3D avatars of participants, known as Digital Twins, are utilized to greatly enhance immersion in the virtual environment. This enhanced sense of feeling and embodiment is essential for successful rehabilitation. The BCI system enables the direct translation of imagined movements into virtual actions performed by the 3D avatar in real time, providing instant feedback to the participant. This feedback is essential for strengthening the neural pathways involved in motor control and recovery. The ultimate goal of the developed system is to significantly enhance the BCI accuracy and effectiveness of MI by making them more interactive and responsive to the cognitive processes of the participants.
A new Chinese gastrointestinal tumor visual language medical model is proposed. This model integrates text and image analysis capabilities, allowing it to simultaneously explain gastrointestinal tumor images and text....
详细信息
ISBN:
(数字)9798350376258
ISBN:
(纸本)9798350376265
A new Chinese gastrointestinal tumor visual language medical model is proposed. This model integrates text and image analysis capabilities, allowing it to simultaneously explain gastrointestinal tumor images and text. The model adopts a framework design that combines a visual transformer and a foundation language model in two parts. It enhances the abilities of feature matching and instruction learning in two stages through deep training. This method improves the model's ability to perform medical summarization and answer complex medical questions. In addition, a gastrointestinal tumor medical image dataset has been organized. It contains more than 20,000 medical images paired with corresponding text. These medical resources have been organized to facilitate comprehensive interpretation of gastrointestinal tumors, which is conducive to assisting gastrointestinal tumor detection and identification. In summary, the introduction of this model and dataset can greatly promote Chinese gastrointestinal tumor detection and recognition, and provide help for many medical workers and patients.
In the field of medical image segmentation, the Ushape architecture model has been widely explored. Recently, a novel architecture called Mamba, based on the State Space Models (SSMs), has emerged. Subsequently, a Vis...
详细信息
ISBN:
(数字)9798331531904
ISBN:
(纸本)9798331531911
In the field of medical image segmentation, the Ushape architecture model has been widely explored. Recently, a novel architecture called Mamba, based on the State Space Models (SSMs), has emerged. Subsequently, a Visual State Space (VSS) block based on visual Mamba has also become a new core module of Ushape architecture models represented by VM-UNet. Although VSS block possesses linear computational complexity, it constructs only four directional linear sequences when processing images, thus inherently lacking the capability to capture local features of the target image effectively. To enhance the capability of VM-UNet in capturing image details, this paper introduces a GM-UNet model, which integrates VM-UNet into the framework of Generative Adversarial Networks (GANs): using VM-UNet as the generator network for image segmentation and employing the encoder of U-Net as the discriminator network. By incorporating the image features extracted by the discriminator and introducing weighted adversarial loss into the network training, the network is trained until the discriminator struggles to distinguish the source of segmented images, thereby obtaining an image segmentation network with superior capability in capturing image details. Experimental results on segmenting liver CT images demonstrate that the GM-UNet model outperforms VM-UNet in various performance metrics and exhibits strong competitiveness compared to other classic segmentation models based on Ushape architecture.
There are several factors that affect the sEMG signal during the process from its generation to its acquisition by sEMG devices. In this study, we tried to explain the physiological functional relationship between sEM...
详细信息
暂无评论