Rapidly assessing intravascular volume
One of the central thrusts of my Ph.D. level research was developing and validating an impedance plethysmographic technique of assessing intravascular volume status (the amount of blood a person has available to work with). To this point, volume status has proven an elusive metric for most currently available techniques, sporting predictive powers between total guessing (AUROC = 0.54) and educated guessing (AUROC = 0.84). The technique I investigated along with researchers from the Department of Emergency Medicine at the University of Michigan proved to correctly assess volume status over 95% of the time over the entire phsyiological range (with many regions of near perfect prediction, AUROC = 1.00).
The technique, referred to internally as dynamic respiratory impedance volume evaluation (DRIVE) used the bioimpedance of the upper limb to predict shifts in blood volume in response to cardiac and respiratory phenomenon. Without going into too much detail (sorry, we are still a little sensitive about IP), we used the respiratory signal to gauge venous return and correlated our results to IVC collapse, a well known predictor of volume status.
Moreover I have developed bioimpedance based techniques to assess intracranial pressure, mean arterial pressure, stroke volume variation, lung recruitment during mechanical ventilation, vasoconstriction, the extent of edema, and correlating mechanical properties to their passive electrical counterparts.
(Almost) All-In-One Medical Wearable
Tracking physiology everywhere it happens
Extending my basic bioimpedance research, I (along with some very talented engineers) designed, created, and validated a wearable physiological sensor capable of measuring heart rate, respiratory rate (and intensity), body temperature, and volume status. From this one device (that’s smaller than a credit card) we were able to measure nearly all the vitals signs a clinician would be interested in.
We started from a basic design, wherein the entire point of the device was to collect the data we needly rapidly and transmit it nearly instanteously to a mobile device for analysis. (Those who are interested in the mobile app and its development are encouraged to consult my labmate TJ’s excellent website for more information.)
Once the basic design was in place, we sought to extend the functionality of the bioimpedance sensor we were using (an AD5933) by creating a reconfigurable frontend that could switch between bipolar and tetrapolar configurations depending on the requirements of the researcher.
We spent a great deal of time and energy on benchtop validations to ensure we were nearly as accurate as a ground truth measurement of impedance from a high-precision LCR meter. It was worth it to us, however, because now we have an inexpensive, wearable medical monitor whose accuracy in human subjects is on par with industry standard equipment.
Advanced IVC Ultrasound Imaging
Taking the guesswork out of evaluation
Using established and novel image processing techniques (Harris corner thresholding, KLT feature-tracking, pyramidal segmentation, etc.) I developed algorithms for automating inferior vena cava (IVC) measurements via subcostal ultrasound. In this way the dynamic interplay of the respiratory and cardiac cycles as a function of volume status could be studied rigorously.
Conducting this research led to the discovery of new physiological phenomenon including detecting a respiratory induced variability to the cardiac signal present in the IVCs motion. This work is currently under review and I hope to see it published here soon.
Dialysis fistula optimization
Moving beyond the coin flip
In many cases arteriovenous fistulas represent a literal lifeline for patient undergoing hemodialysis.H And though they represent a marked improvement for long-term vascular access (compared to grafts and catheters), their exceedingly high failure rate (north of 50% in many cases) demands that we find ways to predict and prevent this failure.
My research monitored three dozen patients’ fistulas from the time of their creation to their maturity (or failure). Among the many aspects of this project, I created an ultrasound imaging toolbox that allowed one to read in standard B-mode images (in cineloops stored as DICOMs) and reconstruct the geometry, track distensibility (using many of the same techniques as the IVC tracking), predict elasticity, and measure the wall shear rate.
From this research we discovered a number of new behaviors including trends elasticity and distensibility that correlate with maturing fistulas. We are now extending this research to produce fluid models to predict vascular remodelling even before surgical intervention.
Blood flow simulations
Predicting the effects of surgery on vascular remodeling
Manufacturing techniques to improve orthopedic surgery
Micromachining K-wire tips
Modifying Kirchner wires to reduce heat accumulation
Tracking in atherectomy
High speed optical tracking
That’s right! Micron level displacement over microseconds in two-dimensions showing the exactly deflection of a (tissue-mimicking) vessel wall for a tool spinning over 3,000 times per second
To correct for needle deflection and tissue deformation
Multi-modal imaging materials
Designing tissue mimicking materials
Using mixed manufacturing modes
Novel electrode validation
Ensuring electrodes don’t get in the way of good electrophysiology