ai4medicine_IterAsian

Welcome
ECG Analysis/sampling and detrending

Motivation

Based on data collected from smartwatches, we are interested in analyzing the metrics collected such as the heartrate, total amount of steps, and the sleep-time to determine whether the user had a high likelihood of having contracted disease before the symptoms start or to report an already existing chronic problem. We were provided with data from 30 labelled users with either COVID-19, influenza-b or other diseases and the timestamp of the onset of symptoms for each disease.

Market / Business Model

Our ideal target group are health-conscious elderly people (>age 65) who already own a smartwatch and are interested in detecting diseases like COVID-19 early on. To further tune the model based on the hardware-specifications of newly released smartwatches and to have access to more data, we are aiming to partner up with hospitals, medical professionals and smartwatch producers. For our business model we chose a monthly-based subscription (recurring revenue) and to make the app free exclusively for certain partners.

Challenges

The main challenge here was basically the dataset and how it was structured. For the initial datasets, we were provided with different features to build a model, once the model is ready we noticed that the dataset for test is just composed of timestamp and heart rates, which made us use other methods of processing and feature engineering to make accurate predictions.

Takeouts

We had the chance to learn a lot in these 2 days, it was mostly the first time for each of us dealing with time series data and it was exciting seeing how each of us tried to handle it.