Deep learning for brain tumor segmentation in multimodal MRI images: A review of methods and advances

Image and Vision Computing, Год журнала: 2025, Номер unknown, С. 105463 - 105463

Опубликована: Март 1, 2025

Язык: Английский

---

KIDBA‐Net: A Multi‐Feature Fusion Brain Tumor Segmentation Network Utilizing Kernel Inception Depthwise Convolution and Bi‐Cross Attention

International Journal of Imaging Systems and Technology, Год журнала: 2025, Номер 35(2)

Опубликована: Март 1, 2025

ABSTRACT Automatic brain tumor segmentation technology plays a crucial role in diagnosis, particularly the precise delineation of subregions. It can assist doctors accurately assessing type and location tumors, potentially saving patients' lives. However, highly variable size shape along with their similarity to healthy tissue, pose significant challenges multi‐label This paper proposes network model, KIDBA‐Net, based on an encoder‐decoder architecture, aimed at solving issue pixel‐level classification errors The proposed Kernel Inception Depthwise Block (KIDB) employs multi‐kernel depthwise convolution extract multi‐scale features parallel, capturing feature differences between types mitigate misclassification. To ensure focuses more lesion areas excludes interference irrelevant tissues, this adopts Bi‐Cross Attention as skip connection hub bridge semantic gap layers. Additionally, Dynamic Feature Reconstruction (DFRB) exploits complementary advantages dynamic upsampling operators, effectively aiding model generating high‐resolution prediction maps during decoding phase. surpasses other state‐of‐the‐art methods BraTS2018 BraTS2019 datasets, accuracy smaller overlapping core (TC) enhanced (ET), achieving DSC scores 87.8%, 82.0%, 90.2%, 88.7%, respectively; Hausdorff distances 2.8, 2.7 mm, 2.7, 2.0 mm.

Язык: Английский

---

MEASegNet: 3D U-Net with Multiple Efficient Attention for Segmentation of Brain Tumor Images

Applied Sciences, Год журнала: 2025, Номер 15(7), С. 3791 - 3791

Опубликована: Март 30, 2025

Brain tumors are a type of disease that affects people’s health and have received extensive attention. Accurate segmentation Magnetic Resonance Imaging (MRI) images for brain is essential effective treatment strategies. However, there scope enhancing the accuracy established deep learning approaches, such as 3D U-Net. In pursuit improved precision tumor MRI images, we propose MEASegNet, which incorporates multiple efficient attention mechanisms into U-Net architecture. The encoder employs Parallel Channel Spatial Attention Block (PCSAB), bottleneck layer leverages Reduce Residual Atrous Pyramid Pooling (CRRASPP) attention, decoder Selective Large Receptive Field (SLRFB). Through integration various mechanisms, enhance capacity detailed feature extraction, facilitate interplay among distinct features, ensure retention more comprehensive information. Consequently, this leads to an enhancement in images. conclusion, our experimentation on BraTS2021 dataset yields Dice scores 92.50%, 87.49%, 84.16% Whole Tumor (WT), Core (TC), Enhancing (ET), respectively. These results indicate marked improvement over conventional

Язык: Английский

---

Revolutionizing diabetic retinopathy detection using DB-SCA-UNet with Drop Block-Based Attention Model in deep learning for precise analysis of color retinal images

The European Physical Journal Special Topics, Год журнала: 2024, Номер unknown

Опубликована: Сен. 20, 2024

Язык: Английский

---