Bioinstrumentation Projects

According to a recent global epidemiology study, about 37 Mn people worldwide (i.e. 0.5% of world population) suffer from atrial fibrillation (Afib). To top it off, this number has grown by 33% in the last 20 years and is expected to grow further due to the aging population and better survival rates. 

Apart from antiarrhythmic drugs and cardioversion, the standard of care is catheter ablation mainly using RF or Cryo energy. Cryoablation is gaining more popularity steadily, thanks to studies like the FIRE & ICE trial, but RF still is the most common one. But a huge drawback of RF ablation is its high failure percentages - about 20-50%. A lot of research has been going on to understand the reason behind these relapses and some plausible reasons are known, but none with good certainty. 

We, at Wolf Lab here at Duke, are on the quest to answer this enigma! To help us do so, we're developing an implantable cardiac monitoring system capable of capturing 16-channel, 24-bit data at 2kSPS and having a longevity of ~3 months (non-rechargeable).

The system has 3 main pieces - the implant, the relay, and the computer. My work has been primarily on the implant - developing system requirements, designing the whole hardware and firmware architecture, and working on the proof-of-concept prototype. The implant has 2x ADS 1298s for data capturing, a PIC18 series MCU, an onboard 1MB FRAM, and a CC110L RF transceiver. 

The system is under development, having finished the data acquisition and data storage parts, but you can check out the slide deck to see where we currently are in the process.

When I tore apart 3 existing Pulse Oximeters (a low-cost, a mid-range, and a high-end one) in the market, I was surprised to discover that all of the seemingly similar devices had significantly different analog conditioning part. One used a dedicated  Analog Front-End IC, the next had a completely discrete analog circuitry, while the last one integrated the analog conditioning within the photodiode sensor. But apart from this, all the designs had a microcontroller (MCU) for digital processing and at least one seprate power chip (LDO/Boost). This got me thinking if there was a better and more sophisticated way to design this simple circuitry.

Fast-forward weeks of prototyping, speaking with potential customers, translating customer needs into system requirements, establishing TI's market absence, convincing superiors of a potential business opportunity here, collaborating with several cross-functional teams, and finally coming up with a functional proof-of-concept....The Single-chip Pulse Oximeter Design was born!

This novel design utilizes the full potential of the highly-integrated and programmable analog circuitry (called Special Analog Combos) within the MSP430FR2355 MCU. But what good does that do? Glad you asked!

This design meets  both the customer requirements and provides several commercial advantages over other designs out there, including:

• Wholistic reduction in cost (lower manufacturing costs due to smaller PCB footprint; lower inventory costs due to single-chip design; lower assembly costs due to highly integrated nature)

• Significant reduction in power consumption (yields 4 hours extended battery lives on the standard 2x AAA battery source) which becomes a lucrative marketing point for customers.

All these perks come without any compromise on the performance and accuracy of the functionality as it performs at par with the existing products in the market.

To capture any biosignal successfully, careful circuit design considerations have to be made to reduce the noise and to amplify & filter out the signal of interest. This course project gave me a hands-on experience of just that! The goal was to design a low-cost and portable (battery-operated) 2-channel EEG bioamplifier system.

To start with, we were given a bunch of design inputs and a bunch of components to carefully chose from. After that, one thing led to another - schematic design, circuit simulation, PCB layout, in-house milling, board assembly, verification & validation testing, system characterization, and the next thing I know, I'm standing with a fully-functional board in hand!

And how did my board performed? Hmm, let's see....

• Had 20% lower power consumption than the target spec

• Had a 60% cheaper BOM than what we were allowed

• And the part I'm most proud about - it was the most compact board in the class (10x6.6cm) & additionally had all the traces on a single side (unlike all the other boards). Because of this something extra, the board could be milled a lot faster than the rest which the lab staff really appreciated!

• Successfully hit all the other targets in terms of bandwidth, gain, CMRR, etc.

So, in a nutshell, outperformed the given design specs!

During my internship with Texas Instruments, I wanted to make the most of this opportunity I was really fortunate enough to get. So, I turned towards an upcoming internal TI competition held annually called 'DIY with TI' with the theme being 'Technology for Good'. I convinced 3 fellow interns and we decided to prototype a low-cost gesture recognition circuit. This fits the theme perfectly as a gesture recognition circuit could have ubiquitous application - one being in Operating Rooms. Currently, surgeons in ORs have to rely on a nurse to operate the equipment (such as display screens) since their hands are not sterile. With this technology integrated within the equipment, the surgeons can use their hand gestures to control any associated function (change magnification/ frame/ audio/ brightness).

Then began the hustle! We'd stay after the office to work on this project (sometimes even till midnight towards the end), we'd seek help from our colleagues whenever we got stuck on anything, and used to distribute work amongst us to maximize our efficiency. In the end, we did reach our goal!

Our prototype had all the functioning in the analog domain. This eliminated the need for any digital component bringing the cost down substantially. We also used off-the-shelf IR sensors for detecting hand motion while keeping prices low. Our final design detected 5 different gestures that could, in turn, be linked to a certain feature of the equipment. The design features included:

• Detecting & differentiating between an L-R swipe Vs. R-L swipe

• Detecting & differentiating between a near swipe or far swipe

• Detecting a hand slider (far->near->far) with latch-on feature

• Avoiding false detection of any swipe

We didn't win the competition - we pitted against some of the brightest minds in TI having years of experience in designing electronics. 

But I like to believe that we still won. Out of 60 interns in our batch, only the 4 of us had the courage and the curiosity to participate in this competition. Clubbed with everything I learnt along the way, is what I value and cherish more than the prize!