Inspiration
As stated by the World Health Organisation, mass testing has been shown to be of great importance for the successful mitigation of the COVID-19 pandemic. For this reason, there is an urgent need of testing kits. Furthermore, these tests must be fast and periodical to ensure the safety of medical professionals, and will also be crucial for the timely recovery of ordinary activity levels and the end of strict social distancing, avoiding second contagion waves due to social or seasonal reasons.
However, test availability has been a recurring and increasing problem in the past months due its scarcity and usually costly, time and labour-intensive nature. Reliable and fast tests such as ELISA are not able to detect the early stages of the illness, while conclusive tests such as PCR are slow. These characteristics may deem traditional testing strategies insufficient to satisfy the current global needs for these tests.
What it does
GOTA (diaGnostic Optical-morphomaThematic Automation) is a computer-aided fast optical testing method for COVID-19. This differential diagnosis testing method could reduce drastically (up to 20 minutes, when correctly optimised) the time required to perform such tests, and serve as an economically and labour sustainable alternative to current method.
The process:
Sample collection and treatment
The sample is collected as a nasal/throat swab, as performed for current regular COVID-19 PCR testing. It is then centrifuged and subjected to negative staining.
Acquisition
Images of the sample are acquired via microscopy with a TEM or SEM.
Processing & diagnosis
The resulting micrographs are subjected to a convolutional neural network which classifies the samples into three categories:
Coronaviridae (virus family) virions found
Non-coronaviridae virions found
No virions found
How we built it
We have implemented a convolutional neural network using TensorFlow. We have been able to perform proof of concept tests of our idea based on reduced photographic samples (electron micrographs) from various sources, mainly from the Public Health Image Library (PHIL) belonging to the Centers for Disease Control and Prevention (CDC) and the National Institute of Allergy and Infectious Diseases. These tests have been documented and show promise towards the feasibility of the project.
Challenges we ran into
Finding a sufficient amount of images for AI training, along with solving technical requirements to achieve technological and scientific validation of the application.
Accomplishments that we're proud of
- A fresh, sustainable, fast and scientifically sound alternative to current testing solutions.
- Collaborating with dedicated researchers on well-known institutions such as CSIC (Spanish National Research Council) and Universidad Carlos III de Madrid.
What's next for GOTA
Acquisition of the necessary resources to perform real experimentation of our proposal. Experiment design and permits.
Validation research in three stages: Acquisition validation, Image acquisition of viral samples and Analysis accuracy validation
Deployment: Protocol development, Software automation, User interfacing, Software and algorithms distribution
Built With
TensorFlow Python3.8
Log in or sign up for Devpost to join the conversation.