The future of heart rate variability analysis using artificial intelligence
Why use Artificial Intelligence for HRV analyses?
Accuracy
Research has shown that only roughly 24% of the studies related to heart rate variability (HRV) perform visual inspection and manual correction of interbeat interval artifacts. This is mainly attributed to the time consumption required to remove these artifacts.
Outlier detection
In order to reduce the amount of human labour even more, the visual inspection and manual correction can be replaced with outlier detection of interbeat intervals.After annotating the electrocardiogram (ECG) signal using artificial intelligence
Live Learning
After a medical professional has corrected any outliers in the interbeat intervals, the corrected data can be sent back to the artificial intelligence model in order to train on live data.This will result in the model getting more and more accurate over time, reducing the need for correction
Expected Timeline
Q2 2023
Q4 2023
Data Collection
Artifictial intelligence (AI) models requires a lot of data. Without a broad dataset covering a large and diverse group of patients, it is not possible to train an accurate AI model that can handle the uniqueness of each individual.
Training the model
Once we have gathered enough data, our team of developers will get to fine tuning our AI model. We aim to use multiple machine learning techniques to create a model suitable for accurate annotation of ECG signals.
Q1 2024
Q2 2024
Q2 2024
Q2 2024
HRV Analysis
Artifictial intelligence (AI) models requires a lot of data. Without a broad dataset covering a large and diverse group of patients, it is not possible to train an accurate AI model that can handle the uniqueness of each individual.
Future expansions
Once our platform can perform the HRV analyses for you, we will expand our platform with even more great artificial tools to aid medical professionals do their amazing work.
Q3 2024
Future
We need your help!
Training artificial intelligence models requires a large amount of data. With a too small dataset, our model will not be able to adapt to the unique electrodiagrams of the heart rate of each patient
We are currently are searching for medical professionals or researchers that would like to share data with us. We would prefer a steady flow of data which has been anonymized by you. However, we are happy with any data you can share, even if you can only share data once. If the data you send is not anonymous, we will anonymize the data before storing it. This way there is no risk to sharing your data with us
The data does need to adhere to a few requirements:
- The data must preferrably be raw from the ECG monitor. No preprocessing is required. If you do have preprocessed data we are still happy to receive it. In case of preprocessed data we do need a specification of what preprocessing was applied.
- For each ECG signal we need the make and model number of the ECG monitor used to gather the data. If you only work with a signel ECG monitor make and model type, We can specify this for all data we receive from you. This why you do not need to keep track of this yourself. If you do use multiple ECG monitor makes and model types, please let us know which model was used for each ECG signal you share with us.
- In case of non-anonymized data we need the patients consent before you send us the data. If you anonymize the data yourself, this consent is not required. Please note that we will still anonymize data if you send data containing information which can be used to identify an individual patient.
Ready to share your data? Please contact us so we can setup a data sharing flow that best suits your own work flow. We aim to make the sharing of data as easy as possible to reduce the impact on your team