Revolutionalizing Medical Diagnostics for Tata MD using Machine Learning
How we used cutting edge machine learning and image processing frameworks to transform medical diagnostic workflows for Tata MD
The Challenge
During the onset of COVID-19, Tata Medical and Diagnostics (Tata MD), the new healthcare venture of the Tata group, intended to launch Rapid COVID-19 diagnostic kits. The kit is powered by FELUDA, a CRISPR CAS9 technology that was developed in India by CSIR- IGIB for COVID testing [1] [2]. It is the world’s first CRISPR CAS-9 based diagnostic tool to be launched globally. The tests were intended to help high volume rapid COVID-19 testing across diagnostic laboratories in the country, to analyse patient samples faster and help the consumer to make informed decisions and reduce the stress of waiting in ambiguity.
The challenge faced by Tata MD was that it wasn't possible to reliably deploy trained personnel across diagnostic laboratories across the country who can quickly take the data from the test assay and precisely update the information so that test results can be guaranteed to be accurate.
The Solution
Working closely with the team at Tata MD, we understood the processes in place and the intricacies of the CRISPR based LFA diagnosis process. We proposed a solution that uses a comprehensive blend of Image Processing and Machine Learning - packaged in a conventional mobile application that runs on Android devices that could be easily deployed to hundreds of laboratories in the country.
The lab technicians could simply snap a photo of the LFA, and the rest would automatically be handled by the platform.
This would radically simplify the cumbersome, and intensely time consuming process of analysing the LFA post the test, and uploading it to a database for later steps.
The Implementation
The implementation of this one of a kind (and very custom) system, required significant R&D with a custom data-set that we had to create ourselves, and we iterated through several models trained with thousands of images to increase accuracy. We worked very closely with the team at Tata MD who helped test the model to judge the accuracy of results.
The model was later packaged inside a micro-service that was queried for inferencing results whenever users at laboratories captured images for tests.
The team at Iesoft worked end-to-end on building the platform - starting from UX, to implementing the mobile app, creating the machine learning model, building the backend microservices and packaging/deploying it.
The best part is that all of these were done within a few months.
Results and Benefits
Once it was rolled out, the system was able to handle thousands of tests per day with a very high degree of accuracy.
Conclusion
Through our innovative approach to developing ML models and Application Development, Tata MD experienced a huge win. By leveraging cutting-edge technologies, we not only achieved a faster turnaround time when it was needed the most, but also managed to deliver a fully functional, reliable and robust technology to the hands of our client. The success of this case study showcases the transformative impact of AI-powered applications in medical diagnostics.
