It might seem ridiculously simple, but machine learning requires training. And training requires training data to rehearse. We do not have enough data to train the neural network models. It is already difficult enough for the AI experts to understand what the physician need, to do so without enough samples.
As PCR tests are in short supply around the whole, and the results os such test requires some time, X-Ray or CT (Computer Tomography) are considered as alternatives in the diagnosis. The X-Ray technology, in particular, is widely available worldwide. It might not be a definitive test, but without question, it is swift.
As related by ZDNet, there are many initiatives on that front, starting in China where the disease was first spotted to established startups like Qure.ai, funded by Sequoia Capital. All those initiatives are facing the same roadblock, lack of enough data sets to train the algorithms.
Paisanos Creando, a design studio in Buenos Aires, is putting together Deepvision an App aimed to help physicians and individuals upload images to build such dataset. A major enabler to further boost AI applications not only to the current but future pandemics.
If specialists upload enough x-ray images of patients infected with COVID-19 and healthy patients, we can further train and improve our AI algorithm and make it available to medical institutions and governments to assist and speed up the process of detecting the virus.
- Upload x-ray images
- Associate images to patient symptoms
- Receive AI algorithm feedback
- Contribute to create a trusted data base
The team just released the initiative after his first COVID-19 web application, InfoCovid, shown significant traction, making it to the leading newspaper in Argentina, La Nación. The web app now available in multiple languages, including English, Spanish and Italian, helps connect demand and supply collaboratively. So the next time you are out and about for toilet paper or gloves, you know where to go.