The problem Arrhythmic.ai solves

Our Machine Learning model focuses on analyzing Electro-Cardiogram tests to detect cardiac arrhythmias. The average human resting heart rate is 60-100 BPM. It is influenced by a lot of factors such as lifestyle, food, etc. Arrhythmia is a condition when this heart rate would either increase above 100bpm or decrease below 60 bpm. The fluctuations in heart rate would be observed in the ECG scans. A distinct pattern can be observed in ECG scans of Arrhythmia patients which is detected by our model. While Arrhythmia is a disorder on its own, it is also a symptom of multiple heart disorders. Our algorithm would be able to recognise the defected heart rate from an ECG scan. It would serve as a preliminary test. Since it requires an ECG, it can only be used in clinics and healthcare centers but would make consultations more efficient, because it provides instant results which would help sort through scans.

Challenges we ran into

The dataset we worked with is unique and preprocessing the vast amount of content to get adequate accuracy was an issue. After a lot of research and trial and error, we finally settled with minimal preprocessing and trained the model. While considering this project on a large scale, we faced certain challenges. Since it involves scanning and sorting through databases for detecting patterns, it requires hardware resources for the training period of the model. To make this a viable product, we have to integrate the model with an ECG and also consider clinical trials. As of now, we have not integrated the model with an application. We are considering making a companion app for the ECG that would give instant test results or figure out a way to directly integrate our model with the ECG.