covidscan.ai

covidscan.ai provides proprietary software that utilizes machine learning-based classification on extracted features from a cxr (frontal chest x-ray). it uses vgg16 as a head model and a customized deep convolutional neural network to analyze covid-19 and pneumonia patients' cxr images. while only being trained on a few thousand images, it still achieves a sensitivity rate of 97.2% and an accuracy between 96% and 97%. more specifically it can detect covid-19 at a rate of 98%, viral infection, and pneumonia without covid-19 at a rate of 98%, and normal x-rays at a rate of 95%. moreover, the software provides visual and text breakdowns of its diagnosis in order to help with severity mapping and treatment. furthermore, cxrs are fast, readily available, and more convenient than using something like a ct scanning machine. this allows for the software to be particularly useful in less resourceful areas where there's a lack of testing materials and a large influx of patients. since our softwa