When it comes to the limitations of our design, we start with the physical device itself:

We intent to use stainless steel, dry electrodes to create a connection to read the electrical signals from the heart to produce an ECG. The problem with this is that the dry electrodes don’t have perfect connections to the skin all the time and can lead to inconsistencies in data collection and lead to discrepancies and holes in the ECG.

The Machine Learning and AI Algorithm also has its own set of limitations:

There are a number of cases that target only one group of people, so it is hard to gather data for a very specific group of people or for more rare diseases and conditions, which leads to inaccuracy and a phenomenon called overestimation or underestimation where in an attempt to increase accuracy the model will fit the parameters exactly to the data in which further errors start to appear. We are also limited on the processing and computing power of devices such as phones and computers because the amount of computations that are done to analyze new ECGs from patients. The last main limitation is that the AI Algorithms are never perfect and always have some very small margin of error, no matter how much data the system has been given to train with.