Blood vessel segmentation algorithms — Review of methods, datasets and evaluation metrics
Introduction
Blood vessel analysis plays a fundamental role in different clinical fields, such as laryngology, oncology [1], ophthalmology [2], and neurosurgery [3], [4], [5], [6], both for diagnosis, treatment planning and execution, and for treatment outcome evaluation and follow up.
The importance of vessel analysis is supported by the constant introduction in clinical practice of new medical technologies aimed at enhancing the visualization of vessels, as endoscopy in Narrow Band Imaging (NBI) [7] and cone beam Computed Tomography (CT) 3D Digital Subtraction Angiography (DSA) [8]. At the same time, standard techniques, such as Magnetic Resonance Angiography (MRA) and Computed Tomography Angiography (CTA), are constantly improved to enhance vascular tree visualization [9], [10], [11].
Manual segmentation of blood vessels is an expensive procedure in terms of time and lacking intra- and inter-operator repeatability and reproducibility. On the other hand, semi-automatic or automatic vessel segmentation methods require at least one expert clinician to segment or to evaluate the segmentation results obtained. In addition, support for the development and evaluation of such algorithms is still poor as publicly available image datasets with associated Gold Standard (GS) segmentation are currently limited to specific anatomical regions, such as retina [12]. However, automatic or semi-automatic blood vessel segmentation could assist clinicians and, therefore, are topics of great interest in medical research, as demonstrated by the high amount of papers annually published in this field. Indeed, an extensive literature already exists on vessel segmentation and in the past years different reviews on vessel segmentation algorithms have been published, such as [12], [13], [14], [15], [16], [17], [18], [19]. However, due to the strong development in the field, updated reviews are required to analyze and summarize the actual state of the art.
This review aims at analyzing a wide spectrum of the most recent and innovative vessel segmentation techniques found in the literature, reporting on state of the art approaches based on machine learning (Section 5), deformable model (Section 6) and tracking methods (Section 7). Moreover, it reports on the most commonly adopted metrics for the evaluation of segmentation results (Section 3) and identifies the available testing datasets (Section 4).
The goal of this review is to provide comprehensive information for the understanding of existing vessel segmentation algorithms by summarizing their advantages and limitations. Each segmentation approach is first analyzed in the general context of image segmentation and then in the specific context of vessel segmentation. For each segmentation category, papers are discussed, illustrating their benefits and potential disadvantages. In addition, summary tables reporting performance measures are presented for each category. The paper concludes with a discussion on future directions and open issues in the field of vessel segmentation.
A summary of the papers analyzed in this review considering year of publication, anatomical region and imaging technique is reported in Table 1. In addition, Fig. 1 highlights the categories of vessel segmentation algorithms analyzed in the following sections of this paper.
Section snippets
Algorithm workflow
As shown in Fig. 1, in vessel segmentation algorithms the input image first undergoes a pre-processing step, which typically concerns noise suppression, data normalization, contrast enhancement and conversion of color image to grayscale image. Since different imaging modalities produce images characterized by different resolution, noise and contrast, different pre-processing techniques have to be employed. An exhaustive review on pre-processing algorithms is presented in [122].
The core of the
Evaluation metrics
Segmentation performance is commonly evaluated with respect to GS manual segmentation performed by an expert clinician. To attenuate intra-subject variability when performing the manual segmentation, and obtain a truthful GS, a combination of segmentations by multiple experts is usually employed. Different strategies have been proposed to combine the segmentations: for example, a voting rule, often used in practice, selects as GS all voxels where the majority of experts agree the structure to
Evaluation datasets
Phantoms presenting meaningful features of interest with respect to the vascular tree (e.g. intensity profile, thickness, tortuosity) are often considered for the evaluation of segmentation algorithms. Phantoms are easy to control and modify with the goal of understanding how and to which degree the algorithm performance depends on parameter settings. Moreover, by providing a GS, phantoms allow simple algorithm validation and training, which is not always an easy task for real clinical images
Machine learning
There are two main classes of machine learning approaches: unsupervised and supervised. The former finds a model able to describe hidden arrangement of input image-derived features, without any prior knowledge or supervision, while the latter learns a data model from a set of already labeled features, as explained in Sections 5.1 and 5.2, respectively. Since unsupervised learning does not require GS segmentation, it is useful for cases where publicly available GS datasets are not available as
Deformable model
Deformable models consider curves or surfaces (S), defined within the image domain, that can move and deform under the influence of internal (Fint) and external (Fext) forces. The former are designed to keep S smooth during the deformation while the latter attract S toward the vessel boundary. Since S initialization is required to start the deformation process, a robust deformable model should be insensitive to the initial position, as well as in general to noise. Recent efforts in deformable
Tracking
Blood vessel tracking algorithms usually consist in the definition of seed points followed by a growth process guided by image-derived constraints. Seed points can be either manually defined or obtained through vessel enhancement approaches. Tracking approaches are particularly useful to segment connected vascular trees, for which the segmentation can be achieved using a limited number of seed points. Differences between tracking-based methods lie in the definition of the tracking constraints.
Conclusion
This review presented a detailed analysis of a wide spectrum of the most modern vessel segmentation techniques. These included image processing methods based on machine learning, deformable models and tracking approaches.
Vessel segmentation dates back to mid 1900s and a solid and rich literature now exists in this field. However, despite the efforts and the already achieved results, there are still opportunities for improvements.
- •
The absence of a systematic evaluation workflow emerges as a
References (208)
- et al.
Blood vessel segmentation methodologies in retinal images–a survey
Comput. Methods Programs Biomed.
(2012) - et al.
A state of the art review on intima: media thickness (imt) measurement and wall segmentation techniques for carotid ultrasound
Comput. Methods Programs Biomed.
(2010) - et al.
A review of 3D vessel lumen segmentation techniques: models, features and extraction schemes
Med. Image Anal.
(2009) - et al.
Cerebrovascular segmentation from TOF using stochastic models
Med. Image. Anal.
(2006) - et al.
Detection of blood vessels in ophthalmoscope images using MF/ant (matched filter/ant colony) algorithm
Comput. Methods Programs Biomed.
(2009) - et al.
Genetic algorithm matched filter optimization for automated detection of blood vessels from digital retinal images
Comput. Methods Programs Biomed.
(2007) - et al.
HoTPiG: a novel geometrical feature for vessel morphometry and its application to cerebral aneurysm detection
International Conference on Medical Image Computing and Computer-Assisted Intervention
(2015) - et al.
Efficient CNN-CRF network for retinal image segmentation
International Conference on Cognitive Systems and Signal Processing
(2016) - et al.
N 4-Fields: neural network nearest neighbor fields for image transforms
Asian Conference on Computer Vision
(2014) - et al.
Learning fully-connected CRFs for blood vessel segmentation in retinal images
Medical Image Computing and Computer-Assisted Intervention
(2014)
Retinal vessel segmentation using the 2-D Gabor wavelet and supervised classification
Trans. Med. Imaging
Vessel segmentation using implicit model-guided level sets
MICCAI Workshop” 3D Cardiovascular Imaging”, Nice France, 1st of October 2012.
A 3D primary vessel reconstruction framework with serial microscopy images
International Conference on Medical Image Computing and Computer-Assisted Intervention
Active contours driven by multi-feature Gaussian distribution fitting energy with application to vessel segmentation
PloS One
Robust segmentation of cerebral arterial segments by a sequential Monte Carlo method: particle filtering
Comput. Methods ProgramsBiomed.
Sequential Monte Carlo tracking of the marginal artery by multiple cue fusion and random forest regression
Med. Image Anal.
3D segmentation of vessels by incremental implicit polynomial fitting and convex optimization
International Symposium on Biomedical Imaging
Precise lumen segmentation in coronary computed tomography angiography
International MICCAI Workshop on Medical Computer Vision
Angiogenesis in cancer and other diseases
Nature
Molecular pathogenesis of retinal and choroidal vascular diseases
Prog. Retin. Eye Res.
Multi-trajectories automatic planner for StereoElectroEncephaloGraphy (SEEG)
Int. J. Comput. Assist. Radiol. Surg.
Statistical study of parameters for deep brain stimulation automatic preoperative planning of electrodes trajectories
Int. J. Comput. Assist. Radiol. Surg.
Automatic trajectory planner for StereoElectroEncephaloGraphy procedures: a retrospective study
Trans. Biomed. Eng.
Validation of a stereo camera system to quantify brain deformation due to breathing and pulsatility
Med. Phys.
Narrow band imaging in endoscopic evaluation of the larynx
Curr. Opin. Otolaryngol.Head Neck Surg.
Cerebral angiography for multimodal surgical planning in epilepsy surgery: description of a new three-dimensional technique and literature review
World Neurosurg.
Contrast-enhanced Magnetic Resonance Angiography for Preoperative Imaging in DIEP Flap Breast Reconstruction
Breast Reconstruction
Dynamic magnetic resonance angiography provides collateral circulation and hemodynamic information in acute ischemic stroke
Stroke
Computed tomography angiography and perfusion to assess coronary artery stenosis causing perfusion defects by single photon emission computed tomography: the CORE320 study
Eur. Heart J.
Recent advancements in retinal vessel segmentation
J. Med. Syst.
A review of vessel extraction techniques and algorithms
ACM Comput. Surv.
A survey on blood vessel segmentation methods in retinal images
Electronic Design, Computer Networks & Automated Verification, 2015 International Conference on
Geometric Methods for Vessel Visualization and Quantification-a Survey
Geometric Modeling for Scientific Visualization
A review on MR vascular image processing: skeleton versus nonskeleton approaches: part II.
Trans. Inf. Technol. Biomed.
Segmentation algorithm of brain vessel image based on SEM statistical mixture model
International Conference on Fuzzy Systems and Knowledge Discovery
Segmentation of liver, its vessels and lesions from CT images for surgical planning
Biomed. Eng. Online
Vessel segmentation from abdominal magnetic resonance images: adaptive and reconstructive approach
Int. J. Numer. Method Biomed. Eng.
Segmentation of liver portal veins by global optimization
SPIE Medical Imaging
Segmentation of lung vessel trees by global optimization
SPIE Medical Imaging
A New Heuristic Function of Ant Colony System for Retinal Vessel Segmentation
Medical Imaging: Concepts, Methodologies, Tools, and Applications
Retinal vessel segmentation: a comparative study of fuzzy C-means and sum entropy information on phase congruency
Int. J. Adv. Rob. Syst.
Improved multiscale matched filter for retina vessel segmentation using PSO algorithm
Egypt. Inf. J.
Multiscale centerline detection by learning a scale-space distance transform
International Conference on Computer Vision and Pattern Recognition
Dense volume-to-volume vascular boundary detection
International Conference on Medical Image Computing and Computer-Assisted Intervention
Hale: healthy area of lumen estimation for vessel stenosis quantification
International Conference on Medical Image Computing and Computer-Assisted Intervention
Robust shape regression for supervised vessel segmentation and its application to coronary segmentation in CTA
Trans. Med. Imaging
Machine learning based vesselness measurement for coronary artery segmentation in cardiac CT volumes
SPIE Medical Imaging
Back-propagation network and its configuration for blood vessel detection in angiograms
Trans. Neural Netw.
Vessel detection in ultrasound images using deep convolutional neural networks
International Workshop on Large-Scale Annotation of Biomedical Data and Expert Label Synthesis
Structure-aware rank-1 tensor approximation for curvilinear structure tracking using learned hierarchical features
International Conference on Medical Image Computing and Computer-Assisted Intervention
Cited by (417)
A detail-oriented super-2D network for pulmonary artery segmentation
2024, Biomedical Signal Processing and ControlAutomated peripancreatic vessel segmentation and labeling based on iterative trunk growth and weakly supervised mechanism
2024, Artificial Intelligence in MedicineOCTA-500: A retinal dataset for optical coherence tomography angiography study
2024, Medical Image AnalysisFFCM-MRF: An accurate and generalizable cerebrovascular segmentation pipeline for humans and rhesus monkeys based on TOF-MRA
2024, Computers in Biology and MedicineA multi-branched semantic segmentation network based on twisted information sharing pattern for medical images
2024, Computer Methods and Programs in Biomedicine