Virtrum Health

Inspiration

Many of us have suffered from some sort of musculoskeletal disorder before. No one likes getting injured. Getting injured is not fun. It is painful, and even rehab is tedious. 11 million Canadians suffer from musculoskeletal disorders each year, and it costs Canadians $37 billion annually.

In order to prevent, diagnose and rehabilitate musculoskeletal disorders, healthcare professionals often require measures of human movement. However, this has traditionally required complex and expensive hardware or subjective visual assessment. What’s more, with the current healthcare systems around the world, prevention and rehabilitation programs are infeasible and expensive to many of us. This often leads to prolonged treatment time and increased healthcare costs. As the impact of musculoskeletal disorders is expected to increase, a significant population in virtually every community would likely benefit from earlier prevention and better rehabilitation programs.

What it does

Let me introduce you Virtrum Health. Virtrum Health is a cloud-based software platform that allows clinicians to perform precise biomechanical assessment simply by using a smartphone camera. By combining scientific studies, sports medicine protocols, computer vision and machine learning, our platform has the abilities to analyze human movement, identify nuances in movement patterns, and provide clear actionable insights for clinicians with a click of a button.

As the patient performs a standardized series of movements, Virtrum’s mobile applications automatically measure and score mobility, stability, and movement control by real-time marker-less 2D/3D vision analysis. As a result, clinicians will get quick and objective data to make more effective treatment decisions. With our monitoring portal, clinicians able to assess treatment progress remotely and get alerts when our advanced analytics detect someone at-risk.

Through Virtrum Health, biomechanical assessments take only 8-15 minutes and can be administered by anyone. The automated assessment ensures consistent evaluation regardless of who is carrying out the assessment, eliminating the need for any expertise or prior knowledge. It is administered with a mobile device, making it highly portable. The short assessment time makes it possible to assess hundreds of patients in hours. Our platform includes a wide range of automated functional movement assessment protocols and is suitable for many different types of individuals.

How I built it

Obtain the labelled source data from the public research dataset UI-PRMD and collect by ourselves as well.
Ingest the raw data via Azure Data Factory and store it in Azure Data Lake.
Implement the machine learning models (e.g. WrnchAI or Posenet) for pose estimation and store on Azure Cosmos DB.
Deploy the trained model onto a web or app service by using Azure Container Instances.
Users take records of their actions and upload to Azure HDInsight, then the processed data will be fed to trained models to obtain the outputs.
Send feedback to users.