Shortly after starting InsightMedi we identified the value of crowdsourcing a database of medical images. It was obvious that even though we started with the idea of building a social platform for healthcare professionals, we could potentially have a real, more direct impact on the quality of the patient outcomes if we could build a dataset large and diverse enough to train machine learning algorithms.
Our strategy for gathering the content — the medical images — was novel in the healthcare world from the get go. Practitioners and students from all over the globe would provide the images through the app, our direct channel of communication with the InsightMedi community. If we could scale our community and create enough value to promote continuous sharing, we might have an interesting formula for unprecedented, low-cost data growth.
Traditional methods for acquiring medical images are extremely expensive and bureaucratic, making it almost impossible for anyone without a massive wallet to even attempt to do this. We knew we had something in our hands that was worth pursuing.
On December 2015 we held our first (official) annual meeting and we set out our BHAGs for the next 10 t0 30 years. Don’t be fooled by our current size, we’re building InsightMedi to stay. We’re creating a company to innovate and become relevant in the healthcare space.
One of the goals we set is:
“To build the biggest database of medical images in the history of mankind.”
And you may wonder what would be the purpose of this? Why would we want to do such a thing and how would it help us make that impact in the quality of care that I mentioned earlier?
One of the main challenges today for physicians is the increasing number of patients requiring healthcare services. More patients mean more health data to be processed, and a lot more pressure to come up with a diagnosis, always in a short amount of time.
We envision InsightMedi as a tool beyond the social aspect of the current platform. We want it to become an ever learning diagnosis machine with the capacity to advise physicians accurately, in a short amount of time. We want to create an Assisted Diagnosis Service, accessible to everyone from within our platform.
So we started drafting a plan to make this possible within the next 5 years. A feasible strategy to build and process the database at scale, and ship a product that puts our newest developments in the hands of our community as fast as possible.
Benefits of this type of service are clear but the challenges might need some serious examination. We’ve been around long enough to know about things like initial skepticism, data jealousy between institutions, image quality issues, platform limitations, costs of infrastructure, and so on.
On top of all of those mentioned above we have legal implications, IP protection, commercialization rights, multi-organizational agreements, and international regulations.
It was clear that this was a grand-challenge type of task, especially for a small company like InsightMedi. It was also clear that if we wanted to tackle this challenge we needed help.
The main two areas where we are in need of support are:
- Building the database: gathering images to use along with our own, and train the algorithms.
- Processing the database: developing the algorithms to do the image analysis.
It is clear that each area brings its own subset of challenges to tackle, but before we got into the details of this we decided to start contacting Research Centers and Universities, to test the overall idea and gather initial impressions.
The proposal described our approach to solve this puzzle, a collaboration strategy between InsightMedi and those institutions looking to participate, working together to develop algorithms based on Fuzzy Logic and Deep Learning techniques. The goal of those algorithms? To extract and represent knowledge from medical images and thus create a cloud-based diagnosis service.
Initial response could not be better and a key factor was the clear value proposition we were making to everyone with whom we talked.
If you like you can watch us present this idea at the Isaac Newton Institute for Mathematical Sciences.
What we bring to the table
A great partnership starts with a complementary skill set and a group of valuable contributions from each side.
Usually, established institutions would provide resources available at Research Centers (eg. Centre for Mathematical Imaging in Healthcare at the University of Cambridge), workforce (developers/researchers), a database of images, its own institutional relationships (eg. Schools of Medicine and Hospitals), and valuable experience at protecting IP and publishing results of all the work developed.
InsightMedi would provide a stable platform that is currently in the hands of thousands of healthcare professionals around the world, its own database of images, its workforce, and partners like Johns Hopkins University, one of our current investors.
Then the synergy obviously comes from taking advantage of each other’s strengths, under a clear collaboration agreement, to tackle a particular problem at a given time (eg., diagnosis of melanoma, heart disease, lung cancer, kidney stones, and more).
Our fast development cycle allows us to ship new versions of the software created through reliable distribution channels (App Stores), measure its impact, and gather new metrics to keep improving the product. A tool to communicate directly and instantly with the real world. Academy and Industry working together side by side to design, implement, and test a what many times would usually be left unused in the pages of scientific publications.
Repeating the same process with every institution that expresses the intention to collaborate is the key to accelerate and scale the creation, and analysis of the largest database of medical images ever known to man. If you’re wondering what we’re doing these days, that is it.
Wanna help? Get in touch!
Thanks for reading,