We are working on Deep Learning based medical analysis systems (Radiology) together with LifeTec.at.
This first project, the Covid-19-Classifier was done in two weekend-coding sprints.
We intentionally quickly open sourced this project, in order to support our medical staff to fight Covid-19.
A publication will follow.
Why?
We believe, adding another tool (Deep Learning, AI) to the set of potential Covid-19 tests is a worth-while endavour. Currently, as of this writing, globally there are not enough PCR tests or reagants for these tests. In addition, CT allows for minimal contact to patients. Therefore, our tool is meant as a proposal for the medical community for an additional test. Note that this does not obsolete or replace a PCR test.
The importance of teleradiology: with the Covid-19 crisis coming to all parts of the globe, it is especially important to help countries with less chance of thorough PCR testing of their respective populations. Therefore, teleradiology can be a great additional tool for classifying / testing for Covid-19. X-ray machines exist in every part of the world. Our approach will also scale to x-ray pictures. For now, for the first version, we focused on CT scans, due to a higher sensitivity.
Frequently Asked Questions (FAQ)
A recent study by Ai T, Yang Z, Hou H et al. has shown that thorax CT images have a high sensitivity w.r.t to COVID-19. Thus, they can potentially detect COVID-19 quicker than the standard PCR lab test can. However, it is important to note that a negative CT scan image can not exclude COVID-19. We therefore see our approach as yet another tool in the toolbox of medical practitioners.
CT scans can potentially show Covid-19 infections even before the PCR test would raise the alarm.
-- (Dr. Domagoj Javor, LifeTec).
Ai T, Yang Z, Hou H et al. Correlation of chest CT and RT-PCR testing in coronavirus disease 2019 (COVID-19) in China: a report of 1014 cases. Radiology. 2020; (published online Feb 26.) DOI:10.1148/radiol.2020200642