GlycoTrem is Skander Helali and Ghida Dandan’s Senior Design Project at Texas A&M University at Qatar. This post will introduce the project and summarize progress at the end of the Fall 2019 semester.
A followup post concluding the project (Spring 2020) will be published around May.
Members and Acknowledgements
This project is part of QNRF grant NPRP 10-1231-160071 through undergraduate student research/employment with L.PI. Dr. Khalid Qaraqe.
The statements made herein are solely the responsibility of the authors.
Special thanks to Dr. Erchin Serpedin, Dr. Lilia Aljihmani, and Sarah Shahin for accepting to be interviewed as part of the ethnographic study.
“An Aggie does not lie, cheat or steal or tolerate those who do.”Aggie Honor Code, Texas A&M University
Problem Statement & Abstract
Diabetic patients suffer from poorly regulated blood sugar (glycemia) levels. If left untreated, the change in glycemia levels presents severe, and possibly life threatening consequences. In spite of the availability of many sophisticated products that continuously monitor glycemia levels, these solutions are costly. The less expensive solutions are invasive and cumbersome for the patients, especially for hypoglycemia (low blood sugar) detection. The reasons behind this characterization is the need to repeatedly prick the finger (test strips) or the arms (sensors).
Current research has shown a possible correlation between hypoglycemia and extremity tremors, for which reason this document proposes a solution for detection of hypoglycemic events by exploiting those tremors. The solution is based on developing applications for wearables that can be connected to a second application for smartphones. The proposed solution, GlycoTrem, provides a non invasive and cost effective solution to this issue. By providing a long term overview of hypoglycemic events along with a higher sampling rate, trends can be easily established and corrective action taken to reduce their frequency.
The market analysis and the benchmarking study with leading competitor products show a clear niche for GlycoTrem’s low-cost, non-invasive properties that will overcome a plethora of current issues faced by consumers in the diabetes management sector.
Customer Needs Survey & Ethnographic Study Results
The survey was designed to investigate the current diabetes solutions in use by consumers, and the desirable features that they seek when choosing one. The survey questions are listed here:
- Do you have or do you know someone who has diabetes?
- What year were you/they born?
- What type of diabetes do you/they have?
- What is your/their gender?
- Are you/they familiar with what hypoglycemic tremors are?
- Have you/they experienced hypoglycemic tremors before?
- How often do you/they experience hypoglycemic tremors?
- How severe are those hypoglycemic tremors?
- How many times a day do you/they measure your/their blood sugar levels?
- What do you/they use to measure your/their blood sugar levels?
- What do you/they like about this measurement tool?
- What do you/they not like about this tool?
- Approximately how much would these tools cost you/them annually?
- How likely are you/they to consider an alternative blood sugar measuring tool?
- Please rate, on a scale of 1 to 5, how important each of the following features are for a blood sugar measuring tool.
- Do you/they own a smartphone?
- Do you/they own a smartwatch?
The survey yielded data from 206 participants, of which 184 were considered valid after sanitizing and scrubbing invalid entries. The following results and analysis will be using this sample size of 184.
The gender distribution above shows that the survey was able to target both genders effectively. There is a slightly higher number of females (48.94%) who undertook the survey compared to men (46.81%). A small number of participants selected other (4.26%).
The percentage of diabetic patients forms almost half the population at 48.15%, whereas that of participants who know somebody diagnosed with diabetes is 38.89%. The reason behind the inclusion of people who aren’t personally diabetic but know someone who is is due to the presence of younger or older patients who are unable to take the survey themselves. Those who answered that they don’t have or don’t know anyone who has diabetes (12.96%) were redirected to the end of the survey.
The percentage of participants being type 1 diabetic is 76.6%, while that of type 2 diabetic participants is 23.4%. This question was asked in order to see if a correlation between the type of diabetes and the tremors was important; however, upon closer inspection of this survey’s results, no direct correlation was deduced.
Most participants (46.81%) measure only between 1-10 times a day. This is followed by those that measure more than 30 times a day (25.53%). Less participants measure 10-20 times a day and 20-30 times a day (17.02% and 10.64% respectively).
Participants could select more than one option in this section of the survey. The majority of participants (106/184) use Blood Glucose Meters. This is followed by Continuous Glucose Monitors (55/184) and Smartphone applications (48/184). Insulin Pumps are used by the least amount of participants along with “other” solutions. (40/184 and 4/184 respectively).
The participants were asked to rate on a scale from 1 to 10 how likely they are to consider a new tool. As shown on the chart, almost half the population isn’t open to the idea of trying out a new product for diabetes management. Further studying of the data, however, showed that 83.2% of those who chose a rating less than 5 were in fact part of the older population (more than 60 years of age) and the younger population (less than 20 years of age).
When it comes to desired features of any replacement tool, the participants were asked to rate the importance of each feature from 1 to 5.
The two critical features were battery life (4.45/5) and cost (4.23/5). The least important aspect of the product was aesthetics (2.68/5) followed by insulin logging and size (3.26/5 and 3.70/5 respectively).
Participants were asked whether they knew what hypoglycemic tremors are. The 27.66% that didn’t know were presented with a brief definition of this symptom.
The overwhelming majority of survey participants report having experienced tremors at some point (76.60%) with only 23.40% reporting that they didn’t.
The participants were then asked to rate how frequently they experience noticeable tremors. A little more than half (58.33%) mentioned that they don’t notice it as frequently, while quarter of the population mentioned that they do (25%). The remaining 16.67% reported that their tremors are occasional.
The majority of participants (80.56%) report that the severity of their tremors is low. This is followed by those who report that they are moderate (11.11%). Only 8% of participants considered their tremors to be severe.
Participants were asked to select if they owned any smartwatch device. The majority of participants answered no (67%). Only 33% reported having a smartwatch device.
Of those that answered yes to having a smartwatch, 46\% reported that they have a watchOS powered watch (Apple). This is followed by 31% of participants who own a wearOS powered watch (Google). Only 23% of particiapants reported having a Tizen smartwath (Samsung).
The majority of participants were willing (yes and maybe) to purchase a smartwatch (90%) with only 10% of participants saying no.
The overwhelming majority of participants reported owning a smartphone (91%) with only 9% answering no.
The percentages of smartphone brand usage is almost equally distributed among Apple and Android. Of those 168 participants who own a smartphone, 82 (49%) use Apple, and 86 (51%) use Android.
All participants were asked to comment on what they like and dislike about the tools that they use.
Continuous Glucose Monitor (CGM) users reported that the item is really expensive. Even though the sensor is supposed to stay on for two weeks, patients have a hard time keeping it on for that long and have to change it more frequently. Other cons to using this product are that the reader is easy to lose, sensor application is painful, having to calibrate frequently, and no alerts are given when the sugar levels go out of range. On the other hand, patients appreciate that the reader provides them with trends of their blood sugar levels.
Those using insulin pumps complain about bleeding because of it, and that the calibrations are frequent and necessary. Additionally, the pump is uncomfortable and disturbs their sleep. The patients do appreciate not having to manually take insulin, along with receiving alerts when their levels go out of range.
Blood glucose meter users dislike having to prick their finger every time they need to measure, along with the fact that the results are sometimes erroneous and they have to do it again, thus wasting strips. Users also dislike having to carry the case for it which makes it easy to misplace. On the other hand, users have reported that this is the most accurate tool therefore making it indispensable.
Comparison to Existing Solutions
The products will be compared using the specifications outlined in their respective data-sheets and support materials. Due to the early stages of development of GlycoTrem, some assumptions will be made regarding some features of the product.
To maintain objectivity and reduce any subjective cherry-picking, the results of the customer survey will be used to highlight the features that users care about the most and to define their weighting in terms of importance.
Due to the early stage of development of GlycoTrem, a direct comparison to third-party solutions is not feasible. In addition, because these are medical products, testing on humans is not a trivial task. As a result, the products will be compared using the specifications and provided data from the manufacturer rather than field studies or real-life comparisons.
This method of product evaluation does not highlight the faults and defects of products that are encountered over a long time of use – it rather compares the ideal use-case of each product which may not be entirely accurate.
The following primary performance metrics were chosen for the comparison using the results of the customer survey and the ethnographic study video: sampling rate, invasiveness, ease of use, price, and accuracy.
The comparison will also include the differences in modes of action of the three devices along with their “pros” and “cons”.
Analysis and Discussion
As shown from the table above, GlycoTrem provides a supplementary solution that reduces the need for frequent invasive measurements of blood sugar levels. GlycoTrem does not aim to replace existing solutions, it aims to complement them.
It is apparent that GlycoTrem excels in being non-invasive, extremely easy to use, and requiring no user interaction while logging continuously. While our proposed solution does not measure actual blood sugar levels, the ability to alert for hypoglycemic events and to log with high resolution the dips in blood sugar level over extended periods of time will surely prove useful to end users. Furthermore, GlycoTrem’s design does not use consumable expensive components, which makes the product cost-effective and competitive with the other solutions.
The two competing products are able to provide accurate numerical values of blood sugar levels. However, they require expensive dispensable sensors, needles, and glucose test strips that pose ethical, economical, and environmental challenges.
GlycoTrem must abide by certain ethical, environmental, economical, and healthy and safety constraints. This product has negligible societal and political impacts.
Since GlycoTrem handles sensitive patient data that can be traced back to them, this poses an ethical dilemma. This issue may completely avoided, however, by employing encrypting algorithms on all gathered information.
GlycoTrem doesn’t produce bio-hazardous waste as opposed to its competitors, eliminating any sizeable environmental impact.
Health & Safety Constraints
The lithium ion batteries used by the smartwatches are tested and validated by the manufacturers themselves. The lithium batteries chosen for the ring, however, will have to undergo certain safety tests.
Functional Model Analysis
Top-Level Functional Model
Representing a system in its top-level one block form allows to visualize the overall input-output relationship of the system’s design. The following diagram shows the top-level condensed view of the system. The black-box can be split into two functional blocks that can be further broken down into more detailed functions and sub-functions.
The primary input of GlycoTrem’s operation is the accelerometer data (x, y, and z axis coordinates – Element 1). This data is then fed into the “Application Layer (Element 2)” which is an abstraction of both the smartwatch and smartphone applications.
The output of this application layer is then fed into the “Processing Layer (Element 3)”. This layer is similarly an abstraction of the dedicated server and database, the outputs here are then fed back into the application layer for visualization (Element 4). This feedback loop does not alter or control the behavior of the system, it is simply data that the application layer can visualize after it was processed.
Some historical data and other information that is produced by the processing layer is stored in the database as represented by the output (Element 5).
These abstractions will be broken down into smaller and more detailed components later on. This will also highlight the hidden inputs and outputs within the blocks represented above in the top-level view of GlycoTrem’s design.
Detailed Functional Model
The two major blocks that can be broken down into more detailed sub-functions are the second and third elements of the top-level view. In the interest of clarity, these blocks will be detailed separately. They will then be all combined in a comprehensive and detailed functional model representation.
The application layer consists of both the smartphone and smartwatch applications. The user only interacts with these two layers. Their primary purpose is to handle the transmission of unprocessed data, visualizing processed data, and alerting in case of hypoglycemic events.
The raw accelerometer data is the primary input of this layer. This is received by the smartwatch application. The processed data received from the server is relayed to the smartphone application which is responsible for visualizing this information.
Within this block, the smartwatch is responsible for continuously transmitting raw accelerometer data using Bluetooth to the smartphone application. On the other hand, the smartwatch application is responsible for sending this raw data unaltered to the processing layer. Additionally, the smartwatch application is responsible for issuing an alerting signal to the smartwatch when hypoglycemic events are detected and received from the server.
The processing layer consists of the database and dedicated server. These handle the bulk storage and the computationally intensive calculations needed to identify hypoglycemic events.
This layer is responsible for the primary function of GlycoTrem which is the identification of hypoglycemic events. The database server receives the unprocessed accelerometer data and relays it to the dedicated server. The dedicated server performs calculations on this data and relays it back to the database server for long term storage. In addition, the dedicated server outputs the processed data to the application layer as outlined previously. The input of this layer is the unprocessed data, while the outputs are the hypoglycemic events.
The database server is expected to archive and keep a record of historical processed data for each patient. This database is responsible for retrieving this information on-demand if requested from any other layer in GlycoTrem’s design. In addition, the database is tasked with relaying the unprocessed data to the dedicated server through an internal UNIX socket (since both servers are containers on the same physical machine).
By combining the functional blocks defined above, the following over-all model is obtained. The model was rendered in left-to-right and top-down formats for clarity.
As shown above, the overall system outline contains one primary input which is the raw accelerometer data, and two primary outputs which are the hypoglycemic events and the visualized data form of the data from the smartwatch.
Inside the functional blocks, more internal inputs and outputs are clearly defined and their purpose is explained in their relevant sections. These were previously hidden in the black-box representation.
Multiple standards exist that govern the secure handling of healthcare data. In addition, there are technical standards for some technologies used by GlycoTrem such as Bluetooth.
HIPAA and HITECH standards regulate the transmission, handling, and storage of patient healthcare data. Stringent safety precautions are to be set in place to protect the anonymity and confidentiality of this sensitive information in the dedicated server. In addition, the IEEE P1752 Standard for Mobile Health Data is applicable due to the handling of the data through a smartwatch and smartphone application. As the smartwatch and smartphone both make use of Bluetooth Low Energy, the Bluetooth standard also applies – and GlycoTrem’s design ought to be compliant.
The app-to-server communication should be encrypted using a TLS socket, GlycoTrem’s design aims to make use of the latest TLS 1.3 standard for optimal security and safety while maintaining good latency through TLS 0-RTT. Should a Web API be developed for integration with third-party tools, it would comply with the JSON:API specification (v1.0).
Many factors come into play that may hinder achieving GlycoTrem’s goals. From filtering unwanted “noisy” data, to protecting the privacy of the users, as well as matching their preferences, all form constraints on the product’s design.
Physiological tremors may be due to several factors other than hypoglycemia, such as fatigue, consuming too many caffeinated beverages, certain medications, or due to Parkinson’s disease. Singling out the tremors that result solely from hypoglycemic events remain as an active constraint to the project’s design. This constraint therefore leads to the following one: accelerometer quality. Hypoglycemic tremors range from 10 to 14 Hz. Since the sampling rate must be at least twice the maximum frequency, any smartwatch that samples at a frequency starting from 30 Hz should be sufficient.
Upon surveying a sample of 184 people, out of which around 100 were diagnosed with diabetes, it was evident that not all patients experience tremors; in fact, it was mostly the younger population that confirmed experiencing tremors often, and this frequency decreased as the age range increased. A possible hypothesis is that the more the patients are subjected to tremors, the less they notice them over time due to being desensitized. Further testing is necessary in order to reach a concrete conclusion.
Current Progress and Future Milestones
All principles of operations and modules of GlycoTrem have been successfully outlined and detailed at this point of this semester. Furthermore, a benchmarking analysis was performed by comparing GlycoTrem to existing solutions, and it yields promising results. In terms of hardware and software, the preliminary prototype for the ring wearable and associated software for data logging have been completed, in addition to completing the Android Wear logging application for smartwatches. The current progress can be seen as per the following table:
The table below shows a concise schedule for the upcoming semester, showing an overview of all the tasks to be accomplished and their assigned deadlines. The tasks seen are also divided among the team members.
A significant market exists for non invasive, low cost solutions in the healthcare field. GlycoTrem aims to fill this niche for diabetic patients who suffer from hypoglycemia unawareness.
The survey and benchmarking analyses conducted show a great demand for a non-invasive, and low-cost solution. They also show a good product competitiveness given the performance criteria selected compared to the leading alternatives in the market.
By the end of the semester, a clear outline of the system’s functionality, base design, and overall concept have been finalized along with a functional prototype / proof of concept, thus paving the way for the continuation and conclusion of the project in the upcoming semester.
I hope you found this project interesting, I’m eager to share follow-up updates as next semester progresses leading to my graduation at May 2020. The project’s website can be found at glycotrem.com. An easier to read PDF version can also be found there.
Feel free to ask any questions if you have any below!