COVID-19 TESTING - AI APPROACH

COVID-19 disease cause pulmonary infections, which could be observed by CT images of the lungs as well as chest X-ray images. Clinicians have been able to observe and diagnose similar infection such pneumonia using this approach. In fact asymptomatic breast cancer and diabetes testing can currently be done using AI image processing.

A anonymised dataset of chest x ray images of Covid-19 infected patients as well as healthy patients is sampled. A convolutional neural network model is trained to analyse chest x -ray images, detect and label COVID-19 infected cases. Mobile chest X ray machines and technicians are relatively more accessible, scalable, and affordable than PT PCR tests and laboratory facilities. A reliable model to assist in the diagnosis of COVID -19 cases and rank patients from most to least severe, in less than a minute of processing time for results. This assists to get doctors to treat the most urgent cases first.

Compared to the 24 hour waiting period for results in the PT PCR approach, that may lead to the nonnegligent spread of the disease in the meantime.

Our model is unique as compared to similar models that are being tested and developed, in that it is trained on reliable, anonymised data of South African patients, making it localised increasing specificity in case detection. False negative cases are negligible in the proposed approach, and work to improve accuracy and specificity is currently being done.