Abstract:Three-dimensional (3D) medical imaging has been incorporated in routine clinical practice, since the required infrastructure has become increasingly affordable. New algorithms and applications are needed to serve the additional image processing and analysis functions in 3D space. In this work we propose a system for semi-automatic modeling and segmentation of elongated salient and anatomical objects in 3D medical images. The proposed methodology is based on a novel mathematical formalization of a well-known class of geometric primitives, namely generalized cylinders (GCs), which exhibits advantages over the existing parametric definition. Since the anatomical objects have to be modeled by their intersection with the transverse image planes, the proposed methodology includes also a new seeded region growing (SRG) segmentation algorithm for ellipse detection in 2D images, based on a priori shape knowledge. Finally, the resulting GC model is used to initialize an active surface (AS) segmentation method, in order to accurately delineate the required object. In this work we present the proposed algorithms in detail, along with the evaluation of the accuracy of the model-based segmentation by experts. Results show that elongated objects like the aorta and the trachea may be segmented with sensitivity between 90% and 95%. The proposed SRG-ellipse detector requires minimal user-initialization and its executions requires only few seconds for each image slice on an average laptop. The evolution of the AS requires less than one second per iteration for a typical CT image. Comparisons are provided with state of the art semi-automatic medical image processing software, which validate the merit of the proposed work.

A Fast 3D Region Growing Approach for CT Angiography Applications.

Adaptive Decomposition and Shared Weight Volumetric Transformer Blocks for Efficient Patch-Free 3D Medical Image Segmentation.

Segmentation of Cardiac Tissues and Organs for CCTA Images Based on a Deep Learning Model

Fast Organ Segmentation in CT Images with CUDA

LRVRG: a Local Region-Based Variational Region Growing Algorithm for Fast Mandible Segmentation from CBCT Images.

Gradient Based Seeded Region Grow method for CT Angiographic Image Segmentation

A Rapid Searching Calculation of Radiation Dose Distribution Based on the Region Growing Algorithm in 3D for CT-Robot Gamma Knife

A Regional Energy Function Based Approach for Fast Tooth Segmentation from CBCT Images

A Novel Multi-stage 3D Medical Image Segmentation: Methodology and Validation

CUDA Based High Performance Adaptive 3D Voxel Growing for Lung CT Segmentation

Efficient Lower-Limb Segmentation for Large-Scale Volumetric CT by Using Projection View and Voxel Group Attention.

Region Growing Algorithm Combined with Morphology and Skeleton Analysis for Segmenting Airway Tree in CT Images

Rapid vessel segmentation and reconstruction of head and neck angiograms using 3D convolutional neural network

Graph-based region growing for mass-segmentation in digital mammography

Two-pass Region Growing Combined Morphology Algorithm for Segmenting Airway Tree from CT Chest Scans

A new medical segmentation method based on Voronoi diagrams and region growing

Automated Segmentation of Coronary Arteries Based on Statistical Region Growing and Heuristic Decision Method

A fast iteration approach to undersampled cone-beam CT reconstruction.

A novel tool for segmenting 3D medical images based on generalized cylinders and active surfaces

Database-guided simultaneous multi-slice 3d segmentation for volumetric data

Fully-automatic segmentation of coronary artery using growing algorithm