Few days ago, we have created diagnosing model which checks that this pneumonia is caused by COVID-19. Our solution got 90%-95% accuracy of COVID-19 diagnosis based on the x-ray scan only. But we need more datasets. Do you have any data from patient with COVID-19? or Do you want to join to us? contact us.
The purpose of this project is to improve an AI model to check that this pneumonia is caused by COVID-19 and predict probability of death, using x-ray scan or CT scan or other patients’ data (e.g. use of information whether the patient has diabetes or other chronic diseases).
Technical details of our experiments:
Step 0 Train pneumonia model based on kaggle dataset ( https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia )
Step 1 Prepare COVID-19 data (https://github.com/ieee8023/covid-chestxray-dataset) and get others chest x-rays from kaggle.
Step 2 Train & Evaluate
– Resnet152 (95%)
– Resnet50 (92%)
– Vgg16 (90%)
Keras & TF / FAST.AI