Echotomography

Background

Ultrasound imaging is a common and relatively inexpensive scanning technology for visualizing internal structures of the body. They are significantly cheap and more portable than MRIs and they do not expose patients to ionizing radiation like X-rays and CT scans. They do come with their own limitations, such as lower resolution, limited field of view, and they require a probe to be in direct contact with the skin.

Echotomography is an imaging technique that attempts to combine the 2D images from ultrasound scans to reconstruct 3D reconstructions of internal structures. This is an active area of research, and some attempts at commercialization have been made, but the machines are expensive and customized for the application of echotomography. My goal is to build a low-cost echotomography system using off the shelf equipment.

The goal

I will consider this project a complete success if I am able to construct a 3D reconstruction of my heart using the equipment. This will be challenging because the heart is a moving structure which will blur the images, and because the heart is located deep in the chest and obscured by the ribs, I will need multiple views of the heart to reconstruct it. To address the blur, I will need ECG gating which I plan to use my previous project on serial ECG data retrieval to accomplish. I also need the dock for the M-Turbo to be able to receive the video output.

The hardware

I have at my disposal a Sonosite M-Turbo with a C60x probe. This is a portable ultrasound machine and a convex probe that is capable of imaging deep structures within the torse. However, this will be insufficient for the heart because the ribs require some steering around, so I will also need a phased array probe.

I also need to be able to track the probe in 3D space. I'm going to attempt to do this using an Intel Realsense D435i camera and AprilTags. I will attach Apriltags to the probe with fixed positions and use the camera to do pose estimation of the probe in 3D space. This will be critical to the reconstruction, because it will allow me to know the position and orientation of each 2D ultrasound image which I can use to reconstruct the volume by voxelizing the pixels and recombing them into a single volume.

The plan

I have the tracking system roughly working, so I need to CAD and 3D print a mount to attach the tabs to the probe. I also need to set up the video output from the M-Turbo and ensure that all of my data streams are properly timestamped and synced. The I will need to write the reconstruction software. With my current probe, the best strategy for a proof of concept is likely going to be a 3D reconstruction of my liver, which is a large, static structure with good visibility and consistent echotexture. If this works, then I can move onto the heart which will require more sophisticated views and the timing will be less forgiving.