MIST: A Simple, Scalable, and Standardized Framework for 3D Medical Image Segmentation Using Deep Learning
MIST is a new open-source framework designed to standardize and streamline the training, testing, and evaluation of deep learning models for 3D medical image segmentation, addressing the challenge of inconsistent methodology and enabling fair comparison of different approaches.
SpineSegDiff: A Diffusion-Based Model for Semantic Segmentation of Lumbar Spine MRI in Lower Back Pain Patients, Enhanced by Pre-Segmentation with nnU-Net for Accelerated Training
This research introduces SpineSegDiff, a novel diffusion-based model for accurate and efficient segmentation of lumbar spine MRI, demonstrating superior performance in identifying degenerated intervertebral discs, a crucial aspect of low back pain diagnosis and treatment.
TP-UNet: Enhancing Medical Image Segmentation with Temporal Prompts
TP-UNet improves medical image segmentation accuracy by incorporating temporal information through textual prompts and aligning semantic representations between text and image modalities.
Label Sharing Incremental Learning Framework for Independent Multi-Label Segmentation Tasks: A Simplified Approach
This paper proposes a novel "label sharing" framework for training a single multi-channel neural network model to perform multi-label segmentation across multiple medical imaging datasets, achieving comparable performance to individually trained models while being more parameter-efficient and enabling incremental learning of new tasks.
FIAS: A Novel Hybrid CNN-Transformer Architecture for Medical Image Segmentation Addressing Feature Imbalance Using Dynamic Fusion and Mixing Attention
FIAS, a hybrid CNN-Transformer network, effectively addresses feature imbalance in medical image segmentation by dynamically fusing local and global features, leading to improved accuracy in capturing both fine-grained details and large-scale structures.
Assessing the Performance of DINOv2 for Left Atrium Segmentation from MRI Images: A Comparative Study with Traditional Deep Learning Models
Pre-trained on natural images, the DINOv2 vision transformer model demonstrates strong potential for accurate and efficient left atrium segmentation from MRI images, even with limited data, outperforming traditional deep learning models in both fully supervised and few-shot learning scenarios.
Segmentation of Small Cerebral Vessels in Ultra-High Resolution 7T MRI: The SMILE-UHURA Challenge
The SMILE-UHURA challenge addressed the lack of publicly available, annotated datasets for segmenting small cerebral vessels in ultra-high resolution 7T MRI by introducing a meticulously annotated dataset and comparing the performance of 16 submitted deep learning methods against two baselines.
Zero-Shot Performance of Segment Anything Model (SAM) Variants for Bone Segmentation in CT Scans: An Evaluation Study and Preliminary Guidelines for 2D Prompting
While not outperforming specialized models, SAM-family models demonstrate promising zero-shot capability for bone segmentation in CT scans, particularly when using bounding box-based prompting strategies.
MLV$^2$-Net: An Automatic Segmentation Method for Meningeal Lymphatic Vessels in 3D FLAIR MRI Addressing Inter-Rater Variability
This paper introduces MLV$^2$-Net, a novel deep learning method for the automatic segmentation of meningeal lymphatic vessels (MLVs) in 3D FLAIR MRI that addresses the challenge of high inter-rater variability in expert annotations.
MSA$^2$Net: Multi-Scale Adaptive Attention-Guided Network for Medical Image Segmentation
This paper introduces MSA$^2$Net, a novel deep learning architecture for medical image segmentation that leverages multi-scale adaptive attention gates (MASAG) within a hybrid CNN-Transformer framework to effectively capture both local and global contextual information for enhanced accuracy and boundary delineation.
© 2024 by Linnk AI