Beatriz Farola Baratam

Ultrasound (US) imaging enables the evaluation of vascular structures in real time, and it can provide morphological and pathological information during US-guided procedures. Automatic prediction of vascular structure boundaries can help clinicians in locating and measuring target structures more accurately and efficiently. Most existing US segmentation methods use per-pixel classification or regression, which require post-processing to obtain contour coordinates. In this work, we present ACE-Net, a novel approach that directly predicts the contour coordinates for every scanning line (A-line) in US images. ACE-Net combines two main modules: a boundary regression module that predicts the upper and lower coordinates of the target area for each A-line, and an A-line classification module that determines whether an A-line belongs to the target area or not. We evaluated our method on three clinical US datasets using, among others, dice similarity coefficient (DSC) and inference time as performance metrics. Our method outperformed state-of-the-art segmentation methods in inference time while achieving superior or comparable performance in DSC. ACE-Net is publicly available at https://github.com/bfarolabarata/ace-net. Graphical ACE-Net is a deep CNN architecture designed for direct contour segmentation of vascular structures in ultrasound images. It formulates segmentation as a boundary regression process combined with the prediction of vascular structure presence. To achieve this, ACE-Net introduces an A-line-based encoding scheme, representing contour coordinates of each vascular structure as two boundary vectors (upper and lower). The network consists of a backbone module, an A-line feature extractor, a multi-scale fusion module, and two final head encoders, which output boundary coordinates and presence probabilities. By leveraging a presence vector, ACE-Net enhances segmentation accuracy and effectively handles common ultrasound artifacts without requiring pre-processing. Its architecture is specifically designed to address the imbalance between lateral and axial information in ultrasound images, and it enables real-time clinical metric computation.

Despite continuous advances in three-dimensional (3D) fusion imaging, two-dimensional (2D) X-ray-based fluoroscopy is still the gold standard intra-operative image guidance tool in endovascular interventions. The adoption of robotic technology offers the potential to bring intra-operative radiation exposure down to a minimum, or even eliminate it. Non-ionizing approaches, such as Intravascular Ultrasound (IVUS), are progressively explored as standalone or fluoroscopy-adjunct techniques for 3D vasculature reconstruction. We have previously demonstrated the feasibility of real-time 3D Main Vessel (MV) modeling from the fusion of IVUS and EM pose data obtained from sensors embedded at the tip of a robotic catheter. This paper proposes to advance MV modeling towards a comprehensive radiation-free 3D guidance framework by means of intra-operative Side Branch (SB) detection and modeling. Two models are proposed to approximate the geometry of SB vessel ostia: a sphere and a cylinder. An Unscented Kalman Filter (UKF) recursively estimates the state of these models considering the MV model, while the catheters navigates through the vessel. In silico and in vitro validation results show the potential clinical value of the proposed strategy for facilitating safer robotic catheter steering.

Intra-operative local 3D vessel representations have the potential to significantly decrease the use of contrast agents and exposure to ionizing radiation during endovascular procedures, while overcoming the 2D visualization limitation of fluoroscopic guidance. By fusing intravascular ultrasound (IVUS) imaging and electromagnetic (EM) pose sensing in a robotic catheter tip, a real-time local 3D model of the vasculature could be constructed intra-operatively. This letter proposes the use of a cylinder model to approximate the vessel geometry near the catheter tip. An unscented Kalman filter is employed to robustly estimate the cylinder that best fits IVUS and EM data while navigating through the vessel. This forms a radiation-free alternative to conventional radiation-based guidance. Validation on one in silico and two in vitro models showed median estimation errors of cylinder radius of 0.14 mm, 0.42 mm and 0.70 mm; cylinder position of 0.45 mm, 1.07 mm and 0.96 mm; and cylinder orientation of $2.94^\circ$ , $4.60^\circ$ and $3.03^\circ$ showing great potential for helping interventionists preventing harmful interactions between the instrument tip and the vessel wall.