BBI Theory in Real Life Applications – Interview with Labfront Research Lead Dr. Francis Hsu

We sat down with Francis, one of Labfront’s research leads, to talk about his research on beat-to-beat interval as an indicator for health and how his work at Labfront has allowed to him continue breaking down research barriers using statistical physics. 

‍

It’s great to sit down and talk with you! Could you please introduce yourself with your professional and academic background? 

In college, I studied Physics and electro-physics with a research focus on medical physics. It was in my third year that I realized I wanted to pursue applied physics instead of theoretical physics because it was more connected to daily life problems.  I then learned about Professor Peng’s research on physiological signals, which is how I became interested in the topic of signal processing and using statistical physics to analyze data. 

‍

What types of research were you doing with physiological signals? 

My PhD thesis was on analyzing physiological signals through heart rate series. My research was to quantify heart rate signals or R waves, with different diseases. Professor Peng’s theory is that the heart rate variation (HRV) of a healthy person should be in the middle between order and disorder. Our daily lives are full of variety, therefore ideally our heart rate should show variety as well and it should fall in the middle. If our heart rate is in the middle, we can adapt to different situations. If you have too much order or disorder in your heart rate then you cannot adequately adapt to challenging life situations. 

This theory was first proposed in 1986, however, before my research, no one had ever used real data and measurements to plot a graph. This graph and model for analyzing healthiness via heart rate was just a concept. My research quantified, using real data, to prove and realize Professor Peng’s theory. 

‍

That’s very neat, so what you did is take real data to prove Professor Peng’s theory?

Yes! Here is my main result. We can make a comparison between the expectation and the result. Here is the disorder line, the complexity axis, and the disorder axis. I defined the algorithm by quantifying the disorder of physiological signals. Through my research, I created a new complexity algorithm called EOE and a disorder algorithm called AE. 

‍

What types of disorders were you studying? 

We were studying patients with atrial fibrillation and congestive heart failure.  

‍

Have you analyzed and researched other physiological signals?   

In another research project, I applied machine learning with a more complicated algorithm in a small study on brain waves. We analyzed brain waves to quantify the severity of Alzheimer’s disease. Through machine learning, we combined different measurements to identify and separate healthy subjects from those with AD. 

‍

What did you discover in your research? 

The advantage of machine learning is it can help us combine complex algorithms. For each signal, we can derive multiple indexes. In brief, we can derive 20 indexes for one series or signal. Our goal was to combine these multiple indexes to separate the healthy brains from the diseased. It was a small study that had 15 healthy subjects and 15 with AD, 30 participants in total. Without the incorporation of machine learning in my research, we wouldn’t have been able to apply multiple indexes. 

To give you an example of what we were doing, if you wanted to separate people by sex assigned at birth, using height as your only index, your results would not be 100% accurate because it wouldn’t take into account that some females are tall. If we factor in more indexes like weight and running speed we will have a more accurate algorithm to differentiate males from females. The power of machine learning is that it enables us to apply multiple indexes. 

During our research, we discovered that if you only consider one index on brain waves the accuracy was 0.83. If we used 2 indexes the accuracy increased 10%. However, when we applied 5 indexes the accuracy became 100%. It was very powerful. 

‍

You must have been so happy when you got the 100% result! 

Yes, it’s a simple method but improved the accuracy significantly. This is why I enjoy applying physics and math to medical science. My research really has a real-life value. 

Through the Labfront platform and Garmin’s wearable devices, there is an opportunity to collaborate with researchers and collect much larger data.

‍

How do you hope your research is used in the real world?

In the first part of my research with the inverted U curve or heart rate analysis, I only validated the graph with 50 subjects. It was a good preliminary discovery that proved Peng’s theory but it still needs to be validated with more data. This is actually the main reason I wanted to join Labfront. 

Through the Labfront platform and Garmin’s wearable devices, there is an opportunity to collaborate with researchers and collect much larger data. By collaborating with other doctors, we can also have more data from different diseases. One of my long-term goals is to see how the theory can be applied to a wide range of diseases. 

‍

Can you give an example of some diseases it could be used in? 

When it comes to heart health, we usually think of the autonomic nervous system. About 6 months ago we collaborated with a doctor and were focused on a disease called sympathetic hyperactivity. They wanted to see the effectiveness of treating the disease with medicine and an operation to suppress the sympathetic nervous system. For the past several months we have been examining where patients fall within the graph as they recover. We want to see different types of diseases in these outer-regions and after treatment, when they recover, to return to the healthy region. 

‍

Can you tell me a bit about your work at Labfront, what is your role and responsibilities? 

My work can be separated into two parts. The first part is I assist researchers using Labfront by calculating indexes for them. They give us access to participant data and then we provide them with the calculation using the heart rate algorithm. 

The second part is I complete validation studies. The most recent was to study Garmin smartwatches' ability to gather heart rate data. My colleague Han-Ping and I tested Garmin wearable devices in different daily life situations such as sleeping, chatting, walking and hiking. Our conclusion was that the Garmin heart rate is accurate and reliable. Researchers can be comfortable using Garmin to measure heart rate during daily activities. 

‍

‍

A researcher that uses Labfront can focus on participant enrolling and management instead of worrying about data management and data analysis.

‍

What can a researcher gain from using Labfront to do their research? 

By using Labfront researchers can collect physiological data easily. Traditionally it’s been hard to get accurate beat-to-beat interval (BBI) data or the heart rate period series. However, using Garmin devices you can obtain heart rate data and through Labfront we have the ability to manage a lot of data. Using the platform, you can collect data from over 400 participants at one time! Participant data is uploaded to the platform automatically, we then help manage and analyze the data. Our team will analyze that data to generate a standard report that includes all required data such as BBI or HRV.

A researcher that uses Labfront can focus on participant enrolling and management instead of worrying about data management and data analysis. We can save not only their time but also provide a professional standard for data processing.