Abstract:Objective. Cardiovascular disease (CVD) is a group of diseases affecting cardiac and blood vessels, and the short-axis cardiac magnetic resonance (CMR) images are considered the gold standard for diagnosis and assessment of CVD. In CMR images, accurate segmentation of cardiac structures (e.g, left ventricle) assists in the parametric quantification of cardiac function. However, the dynamic beating of the heart renders the location of the heart with respect to other tissues difficult to resolve, and the myocardium and its surrounding tissues are similar in gray scale. This makes it challenging to accurately segment the cardiac images. Our goal is to develop a more accurate CMR image segmentation approach. Approach. In this work, we propose a regional perception and multi-scale feature fusion network (RMFNet) for CMR image segmentation. We design two regional perception modules, window selection Transformer (WST) module and grid extraction Transformer (GET) module. The WST module introduces a window selection block to adaptively select the window of interest to perceive information, and a windowed Transformer block to enhance global information extraction within each feature window. The WST module enhances the network performance by improving the window of interest. The GET module grids the feature maps to decrease the redundant information in the feature maps and enhances the extraction of latent feature information of the network. The RMFNet further introduces a novel multi-scale feature extraction (MsFE) module to improve the ability to retain detailed information. Main results. The RMFNet is validated with experiments on three cardiac datasets. The results show the RMFNet outperforms other advanced methods in overall performance. The RMFNet is further validated for generalizability on a multi-organ dataset. The results show the RMFNet also surpasses other comparison methods. Significance. Accurate medical image segmentation can reduce the stress of radiologists and play an important role in image-guided clinical procedures.

A Multi-Resolution Deep Forest Framework with Hybrid Feature Fusion for CT Whole Heart Segmentation.

MRDFF: A Deep Forest Based Framework for CT Whole Heart Segmentation.

Super-Resolution Based Patch-Free 3D Medical Image Segmentation with Self-Supervised Guidance

A cascaded framework with cross-modality transfer learning for whole heart segmentation

Cardiac MRI Segmentation with Focal Loss Constrained Deep Residual Networks

An Improved Combination of Faster R-CNN and U-Net Network for Accurate Multi-Modality Whole Heart Segmentation

Regional perception and multi-scale feature fusion network for cardiac segmentation

MSMHSA-DeepLab V3+: An Effective Multi-Scale, Multi-Head Self-Attention Network for Dual-Modality Cardiac Medical Image Segmentation

FR-NET: Focal Loss Constrained Deep Residual Networks for Segmentation of Cardiac MRI.

Dilated convolution network with edge fusion block and directional feature maps for cardiac MRI segmentation

Whole Heart and Great Vessel Segmentation in Congenital Heart Disease Using Deep Neural Networks and Graph Matching

Multi-residual 2D network integrating spatial correlation for whole heart segmentation

Automatic Whole Heart Segmentation Using a Two-Stage U-Net Framework and an Adaptive Threshold Window.

A Multi-Task Network For Cardiac Magnetic Resonance Image Segmentation And Classification

Whole Heart Region Segmentation from CT and MRI Images Using a Hybrid Model

A Novel Framework with Weighted Decision Map Based on Convolutional Neural Network for Cardiac MR Segmentation

Anatomical Prior-Based Automatic Segmentation for Cardiac Substructures from Computed Tomography Images.

An Integrated Deep Learning Framework for Joint Segmentation of Blood Pool and Myocardium.

Pseudo-3D Network for Multi-sequence Cardiac MR Segmentation

Accurate Segmentation of Heart Volume in CTA with Landmark-Based Registration and Fully Convolutional Network

An Automatic Cardiac Segmentation Framework based on Multi-sequence MR Image