SeqSeg: Learning Local Segments for Automatic Vascular Model Construction.

Computational modeling of cardiovascular function has become a critical part of diagnosing, treating and understanding cardiovascular disease. Most strategies involve constructing anatomically accurate computer models of cardiovascular structures, which is a multistep, time-consuming process. To improve the model generation process, we herein present SeqSeg (sequential segmentation): a novel deep learning-based automatic tracing and segmentation algorithm for constructing image-based vascular models. SeqSeg leverages local U-Net-based inference to sequentially segment vascular structures from medical image volumes. We tested SeqSeg on CT and MR images of aortic and aortofemoral models and compared the predictions to those of benchmark 2D and 3D global nnU-Net models, which have previously shown excellent accuracy for medical image segmentation. We demonstrate that SeqSeg is able to segment more complete vasculature and is able to generalize to vascular structures not annotated in the training data.

Combining ODR and Blood Vessel Tracking for Artery-Vein Classification and Analysis in Color Fundus Images.

PURPOSE: This study aims to develop a fully automated algorithm for artery-vein (A-V) and arteriole-venule classification and to quantify the effect of hypertension on A-V caliber and tortuosity ratios of nonproliferative diabetic retinopathy (NPDR) patients. METHODS: We combine an optical density ratio (ODR) analysis and blood vessel tracking (BVT) algorithm to classify arteries and veins and arterioles and venules. An enhanced blood vessel map and ODR analysis are used to determine the blood vessel source nodes. The whole vessel map is then tracked beginning from the source nodes and classified as vein (venule) or artery (arteriole) using vessel curvature and angle information. Fifty color fundus images from NPDR patients are used to test the algorithm. Sensitivity, specificity, and accuracy metrics are measured to validate the classification method compared to ground truths. RESULTS: The combined ODR-BVT method demonstrates 97.06% accuracy in identifying blood vessels as vein or artery. Sensitivity and specificity of A-V identification are 97.58%, 97.81%, and 95.89%, 96.68%, respectively. Comparative analysis revealed that the average A-V caliber and tortuosity ratios of NPDR patients with hypertension have 48% and 15.5% decreases, respectively, compared to that of NPDR patients without hypertension. CONCLUSIONS: Automated A-V classification has been achieved by combined ODR-BVT analysis. Quantitative analysis of color fundus images verified robust performance of the A-V classification. Comparative quantification of A-V caliber and tortuosity ratios provided objective biomarkers to differentiate NPDR groups with and without hypertension. TRANSLATIONAL RELEVANCE: Automated A-V classification can facilitate quantitative analysis of retinal vascular distortions due to diabetic retinopathy and other eye conditions and provide increased sensitivity for early detection of eye diseases.

Vascular tree tracking and bifurcation points detection in retinal images using a hierarchical probabilistic model.

BACKGROUND AND OBJECTIVE: Retinal vascular tree extraction plays an important role in computer-aided diagnosis and surgical operations. Junction point detection and classification provide useful information about the structure of the vascular network, facilitating objective analysis of retinal diseases. METHODS: In this study, we present a new machine learning algorithm for joint classification and tracking of retinal blood vessels. Our method is based on a hierarchical probabilistic framework, where the local intensity cross sections are classified as either junction or vessel points. Gaussian basis functions are used for intensity interpolation, and the corresponding linear coefficients are assumed to be samples from class-specific Gamma distributions. Hence, a directed Probabilistic Graphical Model (PGM) is proposed and the hyperparameters are estimated using a Maximum Likelihood (ML) solution based on Laplace approximation. RESULTS: The performance of proposed method is evaluated using precision and recall rates on the REVIEW database. Our experiments show the proposed approach reaches promising results in bifurcation point detection and classification, achieving 88.67% precision and 88.67% recall rates. CONCLUSIONS: This technique results in a classifier with high precision and recall when comparing it with Xu's method.

Blood vessel modeling for interactive simulation of interventional neuroradiology procedures.

Endovascular interventions can benefit from interactive simulation in their training phase but also during pre-operative and intra-operative phases if simulation scenarios are based on patient data. A key feature in this context is the ability to extract, from patient images, models of blood vessels that impede neither the realism nor the performance of simulation. This paper addresses both the segmentation and reconstruction of the vasculature from 3D Rotational Angiography data, and adapted to simulation: An original tracking algorithm is proposed to segment the vessel tree while filtering points extracted at the vessel surface in the vicinity of each point on the centerline; then an automatic procedure is described to reconstruct each local unstructured point set as a skeleton-based implicit surface (blobby model). The output of successively applying both algorithms is a new model of vasculature as a tree of local implicit models. The segmentation algorithm is compared with Multiple Hypothesis Testing (MHT) algorithm (Friman et al., 2010) on patient data, showing its greater ability to track blood vessels. The reconstruction algorithm is evaluated on both synthetic and patient data and demonstrate its ability to fit points with a subvoxel precision. Various tests are also reported where our model is used to simulate catheter navigation in interventional neuroradiology. An excellent realism, and much lower computational costs are reported when compared to triangular mesh surface models.