Multi-Scale Tubular Structure Detection in
Ultrasound Imaging
Abstract
We propose a novel, physics-based method for detecting multi-scale tubular features in
ultrasound images. The detector is based on a Hessian-matrix eigenvalue method, but unlike
previous work, our detector is guided by an optimal model of vessel-like structures with respect
to the ultrasound-image formation process. Our method provides a voxel-wise probability map,
along with estimates of the radii and orientations of the detected tubes. These results can then be
used for further processing, including segmentation and enhanced volume visualization. Most
Hessian-based algorithms, including the well-known Frangi filter, were developed for CTA or
MRA; they implicitly assume symmetry about the vessel centerline. This is not consistent with
ultrasound data. We overcome this limitation by introducing a novel filter that allows multi-scale
estimation both with respect to the vessel’s centerline and with respect to the vessel’s border. We
use manually-segmented ultrasound imagery from 35 patients to show that our method is
superior to standard Hessianbased methods. We evaluate the performance of the proposed
methods based on the sensitivity and specificity like measures, and finally demonstrate further
applicability of our method to vascular ultrasound images of the carotid artery, as well as
ultrasound data for abdominal aortic aneurysms.
Existing System
Ultrasound is one of the most widely used imaging techniques in clinical routine,
especially in case of initial diagnosis. This is mainly due to its wide availability and ease of use.
Applications are wide spread, as specialized ultrasound machines and probes are available for
most fields. ultrasound imaging still exhibits many challenges regarding the segmentation and
characterization of tissue. Various approaches, including snakes and active-shape models, rely
on proper initializations to yield correct segmentations. Moreover, not only vascular procedures
could greatly benefit from a visualization of vessel-like structures based on extracted tubular
features, as this could provide physicians with similar possibilities for diagnosis as current
CT-angiography, without exposing the patient to nephrotoxic contrast agents and X-ray radiation
at the same time. After emitting directed ultrasonic waves via a transducer into the patients’
body, these waves get partially reflected at tissue interfaces, which are characterized by a change
in physical properties such as the tissue density. The reflected waves are then received by the
transducer again, and the final image is eventually reconstructed based on these reflections.
As a consequence, this means that no separate receiver behind the target volume is
necessary (in contrast to CT imaging). Based on the different filter responses, we conclude that
the combined measure gives the best overall performance, as it clearly separates the vessel
points, while suppressing high vesselness values in other regions (false positives). This is also
confirmed by the quantitative results, where the combined filter reached the best results - also
regarding Dice coefficients – indicating its suitability for initializing subsequent segmentation
purposes.
Disadvantages of Existing System

The signal attenuation influences the overall image appearance, resulting in deteriorated
image quality for deeper regions.

The resolution and contrast in axial, lateral and elevational directions is not uniform,
resulting in lower visibility of vertical borders.
Proposed System
The proposed a Hessian-based multi-scale tubular structure detection algorithm adapted to the
imaging properties of carotid (3D) ultrasound. Several adaptations of the filter to take different
ultrasound-acquisition characteristics into account, including direction-based and attenuationbased variations. an adaption is exactly the goal of this work, where we try to show how tubular
structure detection algorithms can be redefined for ultrasound imaging by incorporating
information about the specific characteristics.
In this paper, we extend this approach in several aspects: We introduce a multi-scale
estimation not only with respect to the inner diameter, but also to the outer border ring thickness.
To the best of our knowledge, in all prior work, the outer ring was either not considered at all, or
assumed to have a constant radius. We propose a general framework for designing filter kernels
for arbitrary vascular models. We further incorporate confidence maps as prior information to be
able to compensate for ultrasonic attenuation and artifacts within the structure detection.
Advantages of Proposed System

The detection quality of our method remains clearly higher.

The proposed combined filter provides both higher separation and detection than the
other methods.