An attention-fused architecture for brain tumor diagnosis

Biomedical Signal Processing and Control, Год журнала: 2024, Номер 101, С. 107221 - 107221

Опубликована: Ноя. 20, 2024

Язык: Английский

---

Multi-modal global- and local- feature interaction with attention-based mechanism for diagnosis of Alzheimer’s disease

Biomedical Signal Processing and Control, Год журнала: 2024, Номер 95, С. 106404 - 106404

Опубликована: Май 7, 2024

Язык: Английский

---

DASNet: A Convolutional Neural Network with SE Attention Mechanism for ccRCC Tumor Grading

Interdisciplinary Sciences Computational Life Sciences, Год журнала: 2025, Номер unknown

Опубликована: Март 24, 2025

Язык: Английский

---

Deep learning and radiomics-based approach to meningioma grading: exploring the potential value of peritumoral edema regions

Physics in Medicine and Biology, Год журнала: 2024, Номер 69(10), С. 105002 - 105002

Опубликована: Апрель 9, 2024

To address the challenge of meningioma grading, this study aims to investigate potential value peritumoral edema (PTE) regions and proposes a unique approach that integrates radiomics deep learning techniques.

Язык: Английский

---

ADBNet: An Attention-Guided Deep Broad Convolutional Neural Network for the Classification of Breast Cancer Histopathology Images

IEEE Access, Год журнала: 2024, Номер 12, С. 133784 - 133809

Опубликована: Янв. 1, 2024

Язык: Английский

---

ConjunctiveNet: an improved deep learning-based conjunctive-eyes segmentation and severity detection model

International Journal of Intelligent Computing and Cybernetics, Год журнала: 2024, Номер 17(4), С. 783 - 804

Опубликована: Авг. 16, 2024

Purpose The study aims to enhance the detection and classification of conjunctival eye diseases' severity through development ConjunctiveNet, an innovative deep learning framework. This model incorporates advanced preprocessing techniques utilizes a modified Otsu’s method for improved image segmentation, aiming improve diagnostic accuracy efficiency in healthcare settings. Design/methodology/approach ConjunctiveNet employs convolutional neural network (CNN) enhanced transfer learning. methodology integrates rescaling, normalization, Gaussian blur filtering contrast-limited adaptive histogram equalization (CLAHE) preprocessing. segmentation novel method. framework’s effectiveness is compared against five pretrained CNN architectures including AlexNet, ResNet-50, ResNet-152, VGG-19 DenseNet-201. Findings finds that significantly outperforms existing models detecting various stages conditions. demonstrated superior performance classifying four distinct – initial, moderate, high, severe healthy stage offering reliable tool enhancing screening diagnosis processes ophthalmology. Originality/value represents significant advancement automated diseases, particularly conjunctivitis. Its originality lies integration its comprehensive approach, which collectively capabilities. framework offers substantial value field by improving disease classification, thus aiding better delivery.

Язык: Английский

---

Cross-attention guided loss-based deep dual-branch fusion network for liver tumor classification

Information Fusion, Год журнала: 2024, Номер unknown, С. 102713 - 102713

Опубликована: Сен. 1, 2024

Язык: Английский

---

RELA_Net: Upper Airway CBCT Image Segmentation Model Based on Receptive Field Expansion and Large-Kernel Attention

IEEE Access, Год журнала: 2024, Номер 12, С. 89713 - 89725

Опубликована: Янв. 1, 2024

The structure of the upper airway is variable and complex due to its environmental physiological factors. Currently, doctors mainly rely on manual outlining segmentation from images. This method time-consuming relies heavily doctor's experience. To solve this problem, we propose a fully automatic model for Cone Beam Computed Tomography (CBCT) images based U-Net. receptive field expansion module (RFEM) used replace last three convolutional blocks encoder in original U-Net improve feature information extraction capability. And large kernel attention (LKA) added skip connection part dynamically adjust backbone, alleviate loss redundancy connection. dataset paper one created by us clinicians themselves, totaling 1345 CBCT Which were taken 53 patients with obstruction. imaging experts guided delineated label Experimental results show that IoU Dice score predicted RELA_Net network article test sets are 94.39% 97.10% respectively. Based prediction maps set images, proposed demonstrates an improvement comparison other models, particularly reducing over- under-segmentation airway. contributes improving diagnostic accuracy obstruction, thereby enhancing patient care treatment planning.

Язык: Английский

---

CytoNet: an efficient dual attention based automatic prediction of cancer sub types in cytology studies

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Окт. 28, 2024

Computer-assisted diagnosis (CAD) plays a key role in cancer or screening. Whereas, current CAD performs poorly on whole slide image (WSI) analysis, and thus fails to generalize well. This research aims develop an automatic classification system distinguish between different types of carcinomas. Obtaining rich deep features multi-class while achieving high accuracy is still challenging problem. The detection cancerous cells WSI are quite due the misclassification normal lumps cells. cluttering, occlusion, irregular cell distribution. Researchers past mostly obtained hand-crafted neglecting above-mentioned challenges which led reduction accuracy. To mitigate this problem we proposed efficient dual attention-based network (CytoNet). composed two main modules (i) Efficient-Net (ii) Dual Attention Module (DAM). capable obtaining higher enhancing efficiency as compared existing Convolutional Neural Networks (CNNs). It also useful obtain most generic it has been trained ImageNet. Whereas DAM very robust attention targeted negating background. In way, combination module robust, intrinsic comparable performance. Further, evaluated well-known datasets Our generated thyroid dataset Mendeley Cervical (Hussain Data Brief, 2019) with enhanced performance their counterparts. CytoNet demonstrated 99% rate comparison its counterpart. precision, recall, F1-score values achieved 0.992, 0.985, 0.977, respectively. code implementation available GitHub. https://github.com/naveedilyas/CytoNet-An-Efficient-Dual-Attention-based-Automatic-Prediction-of-Cancer-Sub-types-in-Cytol.

Язык: Английский

---

Quaternion Squeeze and Excitation Networks: Mean, Variance, Skewness, Kurtosis As One Entity

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 174 - 189

Опубликована: Ноя. 29, 2024

Язык: Английский

---