Applying our AI
Our platform
We deploy AI directly on our platform. This means users will automatically benefit from our continous improvements
API for analysis
We provide API and documentation for other platforms that seeks to leverage our AI development
Datasets
We collaborate with researchers to ease their data analysis process. Contact us for more details on how we can help you
Validations
In order to ensure safety of our AI algorithms, we invesitagate and evaluate their performance with the data that was not used in the development process. It doesn’t mater if you are numbers-loving geek or just a curious person: below you will find some more technical papers concerning those issues.
View More
Technology
About Our AI
Medsensio uses Artificial Intelligence algorithms to help you spot the abnormalities in the auscultation recordings.
2016
Tromsø 7
The initial work on our AI was started in 2016 as a collaboration between UiT The Arctic University of Norway and the population study Tromsø Study 7
2019
ML pipeline
We developed a dedicated pipeline for machine learning in order to automate more of the procedure, helping us to speed up model improvements
2020
New functionality
In 2020 we started development of our next algorithm, intended to provide more detailed information about lung health
2022
Certification
We will CE certify our lung sound analysis tool as a medical device by the end of 2022
Our data
In order to develop our algorithms we use data collected in the Tromsø Study, one of the worlds most comprehensive population studies. In addition, we use data obtained from chronic cardiopulmonary patients and Covid-19 patients in the Horizion 2020 project PyXy.AI

Analysis process
It doesn’t matter if you are a medical professional, student or patient - our algorithms need your help only in the recording collection process. The algorithm will analyse the recording and inform you whether it is of sufficient quality and whether any abnormal findings are present

Our algorithms
Our AI uses spectrograms to analyse recordings and recognize abnormal breathing. We display that same spectrogram for you, as research has shown auscultation analysis performance improves when both audio and visual data is combined
