Document Type

Restricted Campus Only

Publication Date

5-6-2022

Abstract

For those in impoverished communities or remote regions, obtaining adequate healthcare can be a burden. Furthermore, limited access to specialists like cardiologists can make curable conditions a death sentence by leading them to be identified too late. An essential factor in the identification of heart conditions is the use of an electrocardiograph to measure the signal of the heart. In this project, the Remote Heart Diagnosis Team endeavored to design and build a prototype capable of remotely collecting and analyzing an electrocardiogram and displaying the results to a cardiologist in any location for review.

The prototype is composed of five main components. The first component is a printed circuit board designed to record the electrocardiogram. The second is a Raspberry Pi and touchscreen programmed to guide the user through the collection process, compile patient data and read the output of the circuit, run an artificial intelligence algorithm, and store all the information in the third component, a remotely deployed database, using a wireless connection. The fourth component is a website that accesses the database and allows doctors to view and interact with the device data. The final component is a three dimensional printed casing that houses the circuit, microcomputer, and touchscreen.

In early stages of testing, the team identified the need to transfer the circuit from a breadboard to a printed circuit board as the circuit often failed after being moved due to loosened wires. The team also discovered that the noise in the circuit was dependent on the wall outlet being used, leading to the addition of a filter in the circuit. As shown in the success of all but one final test, the prototype meets all expected qualifications, allowing for the changes in the potential diagnoses with the approval of the project sponsor. The only design requirement not achieved was in regards to the ability of the website to replicate a commercial electrocardiogram in form with 90% accuracy; however, as there were limitations with the commercial electrocardiogram used in terms of details in the data, measurement methods, and accuracy, the visuals were deemed reasonable due to their similarity to a traditional electrocardiogram.

Overall, the prototype is a fully functional proof of concept, as it is able to measure a clean electrocardiogram signal from a patient and collect their biographical data, analyze the electrocardiogram using a deployed artificial intelligence network, and store the results in a manner that can be remotely accessed on a website. This prototype is only a proof of concept, however, as, while the developed artificial intelligence network proved that the deployment and use of this type of network is possible, the network is unable to achieve functional accuracy due to a limited dataset. Before commercialization of the prototype, the neural network would need to be retrained using a large dataset of electrocardiograms collected using the device and labeled by a trained cardiologist. Furthermore, the neural network architecture and its ramifications in terms of classification should be reviewed with a professional cardiologist to ensure that image classification has the potential for functional accuracy. In addition, while outside the scope of this project, security code would need to be added to protect the website and transmissions before the prototype could be commercialized to comply with patient privacy laws and medical information regulations. Therefore, while the prototype was successful in achieving the desired functionality and in meeting the requirements, additional improvements will need to be made prior to the expansion of use.

Comments

Prof. Mehran Aminian, Project Sponsor/Team Advisor

Dr. Keith Bartels, Technical Advisor

Dr. Darin George, Senior Design Administrator

Download

Share

COinS