BioPhotonics spoke with Andre Faubert, a biomedical engineering undergraduate at the Stevens Institute of Technology. He has worked at Shang Wang’s Biophotonics Lab since May 2021. His research with Wang surrounding 4D optical coherence tomography (OCT) and embryonic mouse heart development was recently published in a paper (www.doi.org/10.1364/boe.475027).

4D OCT imaging establishes a functional relationship between blood flow and heart wall dynamics in early embryonic development. How will this potentially inform clinicians about the onset of heart defects?

4D OCT imaging is essentially a 3D video showing one complete heartbeat. By identifying the heart rhythm and using it to synchronize a 3D scan, we can visualize dynamic processes such as blood flow and heart wall motion in 3D as the heart beats. In our current research, we quantify flow parameters, such as the timing and speed of the contraction wave brought about by the electrical activity of the heart, along with the amount of retrograde (backward) flow, which is known to affect the development of the heart on a cellular level as the cells respond to the shear stress induced on the inner lining of the heart. We suspect these play a key role in the mechanism of proper valve and septum formation. Improper valve formation can result in valve regurgitation, where a seal fails to form and retrograde flow continues into adulthood, lowering cardiac efficiency. Similarly, an atrial septal defect — for example, a hole in the wall separating the atria, or upper chambers of the heart — can cause oxygenated blood in the left atrium from the lungs to regress into the right atrium, reducing the efficiency of both the heart and lungs.

Since ours is primary research, it is difficult to gauge what impact we may have on clinical practice. At the least, we will better inform clinicians as to the underlying causes of congenital heart defects. Better, we might provide more information as to whether a hole might close by itself. At best, we could inform the development of drugs that remedy the root cell-level cause of the defect, perhaps even causing a septal defect to close.

Does the ability to rearrange images corresponding to the heartbeat enable the high-speed collection of images in both space and time?

Yes. Ordinarily, you can’t afford to spend much time on an acquisition because the movement of the heart distorts the image. The typical method involves acquiring slices so fast that the heart has not significantly moved before a volume acquisition completes. This traditional method requires a reduced spatial sampling rate to satisfy the data rate limits inherent to all image acquisition systems. However, the heart occupies a niche case: Because the heart beats periodically, we can dwell on each slice for one complete heartbeat cycle with the 4D OCT method before advancing to acquire the next cycle at the next location. This periodicity allows us to spend the time needed to achieve high-resolution images. We can spend more time acquiring more data by reordering how the information is collected.

Your paper indicated that you used a custom-built spectral domain OCT system to acquire your results, but it also indicated that your synchronization method was open source. How difficult would it be for other laboratories, or perhaps clinics, to duplicate the creation of time-difference cardiograms?

The beauty of this technique is its simplicity and ubiquity. The algorithm is designed to transform a 3D image into a 4D image by rearranging individual 2D images (B-scans) such that those that share a phase — at the same time in the heartbeat — go together to form a volume. As such, the scan pattern represents an ordinary volume acquisition, except with the B-scans set to sample at high density. Our imaging setup is custom-made but quite typical. Every off-the-shelf OCT system that can perform 3D acquisitions can also perform the simple scan pattern we use. The speed and simplicity of the algorithm should allow anyone with a MATLAB subscription to download our algorithm with the provided example data set from GitHub and get started performing their own 4D alignments. The computational requirements are also minimal. Additionally, the algorithm is nonspecific to the imaging modality. Although we have not tested it as such, I can’t think of any reason why it wouldn’t work with any volumetric imaging technique, not just for OCT.