Pipeline Info

Good Morning,

we had the following question about the covid pipeline: we give about 400 dicom images of a patient to the system which gives the total result present in the “/clara/payloads/<PAYLOAD_ID>/operators/classification-covid-19/classification/preds_model.csv” file. Is the percentage (COVID / non-COVID) relative to the number of images provided? That is, for example, a percentage of 80% COVID on a set of 400 images of a patient, does it mean that COVID was found in 320 images? Or is it a level of confidence obtained in another way?

Thank you in advance for your response.

Best Regards

The classification inferences on the whole lung, which is first segmented out by the lung segmentation model. The final outcome gives the probability of COVID-19 infection. So, in theory all slices are taken into consideration as far as the pipeline is concerned.