Sparse Ultrasound Image Reconstruction From a Shape-Sensing Single-Element Forward-Looking Catheter

Minimally invasive procedures, such as intravascular and intracardiac interventions, may benefit from guidance with forward-looking (FL) ultrasound. In this work, we investigate FL ultrasound imaging using a single-element transducer integrated in a steerable catheter, together with an optical shape sensing (OSS) system.

We tested the feasibility of the proposed device by imaging the surface of a tissue-mimicking (TM) phantom and an ex vivo human carotid plaque. While manually steering the catheter tip, ultrasound A-lines are acquired at 60 Hz together with the catheter shape from the OSS system, resulting in a two-dimensional sparse and irregularly sampled data set. We implemented an adaptive Normalized Convolution (NC) algorithm to interpolate the sparse data set by applying an anisotropic Gaussian kernel that is rotated according to the local direction of the catheter scanning pattern. To choose the Gaussian widths tangential ( ${\sigma _t}$) and normal ( ${\sigma _n}$) to the scanning pattern, an exhaustive search was implemented based on RMSE computation on simulated data.

Simulations showed that the sparse data set contains only 5% of the original information. The chosen widths, ${\sigma _n} = \text{250}\;\mu {\textrm{m}}$ and ${\sigma _t} = \text{100}\;\mu{\textrm{m}}$, are used to successfully reconstruct the surface of the phantom with a contrast ratio of 0.9. The same kernel is applied successfully to the carotid plaque data.

The proposed approach enables FL imaging with a single ultrasound element, mounted on a steerable device.

This principle may find application in a variety of image-guided interventions, such as chronic total occlusion (CTO) recanalization.