MRD-UNet: a compute-efficient medical image segmentation architecture with multi-attention and parallel dilated fusion

AMDF-UNet: a boundary-enhanced adaptive multi-scale feature fusion network for abdominal CT multi-organ segmentation

Accurate segmentation of abdominal CT images plays a critical role in clinical diagnosis and surgical planning. However, the significant size variations among abdominal anatomical structures — ranging from large organs such as the spleen and spine to small lesions such as urinary stones — together with blurred boundaries between adjacent organs, pose substantial challenges for automated segmentation methods. To address these issues, we propose AMDF-UNet, a novel segmentation network that integrates a Boundary-Enhanced Channel-Prior Convolutional Attention (BE-CPCA) module and an Adaptive Multi-scale Dilated Fusion (AMDF) module into the U-shaped encoder-decoder architecture. The BE-CPCA module is deployed at the bottleneck layer, combining dual-pathway channel attention with a Sobel-operator-based boundary-enhanced spatial attention mechanism to simultaneously capture global channel dependencies and fine-grained boundary features. The AMDF module is embedded in the decoder, employing parallel dilated convolutions with learnable adaptive weights to fuse multi-scale contextual information, thereby enabling effective segmentation of structures with diverse scales. The overall training objective combines Focal Loss, Dice Loss, and a morphology-based Boundary Loss to jointly optimize pixel-level classification, region-level overlap, and boundary-level accuracy. Experiments on a clinical abdominal CT dataset comprising five anatomical categories demonstrate that AMDF-UNet achieves a mean Dice Coefficient of 93.45% and a mean IoU of 92.13%, outperforming mainstream methods including UNet, UNet++, DeepLabV3+, Attention U-Net, ResUNet, EGE-UNet, TransUNet, Swin-UNet, and VM-UNet.

Attention-Enhanced Bimodal 3D Medical Image Segmentation with Two-Stage Learning

Computer-aided diagnostic technologies have demonstrated substantial advantages in 3D medical image segmentation, particularly in multimodal 3D medical image segmentation tasks, where they play a pivotal role in driving continuous innovation in related architectures. As an integration of U-Net and Transformer, the UNETR architecture has demonstrated remarkable efficacy in 3D medical image segmentation. Nevertheless, despite its successes, UNETR remains challenged by clinical complexities such as intricate tumor localization and anatomical structural diversity in complex clinical settings. To address these issues, we propose an enhanced 3D segmentation framework, UAtten-Unetr, designed to improve segmentation accuracy and robustness in complex medical scenarios. The framework captures global contextual information via hierarchical Transformer layers and incorporates a spatial–channel attention module to enable adaptive fusion of multimodal features, thereby effectively enhancing cross-modal feature alignment capabilities. Concurrently, we innovatively developed a unified loss function based on bimodal modality-specific Dice constraints and uncertainty regularization, optimized for synchronous learning across the ACDC (cardiac MRI) and AMOS22 (abdominal CT/MRI) datasets. Experimental results showed that UAtten-Unetr achieved an average Dice score of 92.20% on the ACDC dataset, exceeding the reported nnU-Net result of 91.61% by 0.59 percentage points. On the AMOS22 dataset, the proposed method achieved an average Dice score of 84.51%, exceeding the reported UNETR result of 78.33% by 6.18 percentage points. However, its myocardium Dice score (84.11%) was lower than those of nnU-Net (89.24%) and MT-UNet (89.04%), indicating a remaining limitation in myocardium boundary segmentation. These results indicate competitive segmentation performance under the reported experimental settings. This method delivers dual improvements in accuracy and generalization across complex anatomical scenarios, providing an effective solution for precise diagnosis in intricate clinical environments.

DSGH-Net: Medical Image Segmentation via Dual-Statistical Dynamic Context and Graph-Convolutional Heterogeneous Decoder.

A Cascaded Deep Learning Framework for Robust Liver CT Segmentation Using ROI Refinement and Patient-Level Cross-Validation

Accurate liver segmentation from computed tomography (CT) images is essential for computer-aided diagnosis, treatment planning, volumetric assessment, and longitudinal monitoring of patients with hepatocellular carcinoma. However, automated segmentation remains challenging because of variations in liver morphology, low-contrast boundaries, pathological abnormalities, and neighbouring abdominal structures with similar intensity characteristics. This study presents a failure-aware cascaded deep learning framework for automated liver CT segmentation using the publicly available HCC-TACE-Seg dataset. The proposed approach first performs coarse liver localisation on the complete CT image and then extracts a region of interest with an additional safety margin. The cropped region is subsequently processed by a high-resolution U-Net++ model with a ResNet34 encoder. A hybrid Dice and binary cross-entropy loss is used during training, while validation-optimised thresholding, largest-component filtering, and morphological hole filling are applied to improve anatomical consistency. Monte Carlo dropout is further incorporated to estimate predictive uncertainty and support case-level quality assessment. Following annotation verification, 103 liver-positive representative CT images were retained after excluding one image with an empty liver mask. Five-fold cross-validation was performed at the available patient-identifier level to prevent samples from the same patient from being distributed across training and testing partitions. The proposed framework achieved a mean Dice similarity coefficient of 0.8171, Intersection over Union of 0.7442, precision of 0.7993, sensitivity of 0.8737, and specificity of 0.9750. Boundary-based evaluation yielded an HD95 of 48.91 pixels, an average symmetric surface distance of 13.41 pixels, and a surface Dice score of 0.4250. Probability calibration produced a Brier score of 0.0418 and an expected calibration error of 0.0811. Uncertainty estimates were additionally used to flag potentially unreliable segmentations for manual review. These findings indicate that cascaded localisation and region-of-interest refinement can provide robust liver segmentation while reducing background interference and supporting uncertainty-informed quality control.

Design of U-Net Architectures for Medical Image Segmentation using AI/ML Model

Fully convolutional encoder-decoder networks, in particular U-Net, are the leading method for dense pixel-level classification among the deep learning techniques and do not have any black-box problem and that makes the faster and more accurate in segmentation.

BE-Unet: a boundary-enhanced UNet for skin lesion segmentation

Automated skin lesion segmentation is critical for computer-aided diagnosis (CAD) of dermatological diseases. While UNet and its variants are widely adopted, they face challenges in extracting robust contextual features and aggregating discriminative information due to intrinsic intra-class variation, inter-class similarity, and noise interference in medical images. To address these limitations, we propose BE-Unet, a novel Boundary-Enhanced UNet architecture. BE-Unet integrates three key innovations: Large Paralleling Kernel Attention (LPKA) enhances focus on critical features prone to degradation during early encoding and late decoding stages. Dynamic Weighted Group Multi-axis Hadamard Attention (DHA) augments global-local feature perception. Boundary Enhancement Module (BEM) explicitly addresses ambiguous boundaries via dual-task learning, comprising: (1) Segmentation-Boundary Generation (SBG): Simultaneously predicts lesion masks and boundary maps. (2) Feature Enhancement Fusion (FEF): Dynamically fuses multi-scale features using segmentation and boundary guidance. Extensive experiments on ISIC2017 and ISIC2018 datasets demonstrate that BE-Unet significantly outperforms state-of-the-art methods in segmentation accuracy and boundary delineation, with a parameters count limited to 38 KB and Giga-Operations Per Second (GFLOPs) limited to 0.1.