Multi-task Deep Learning Based CT Imaging Analysis For COVID-19: Classification and Segmentation

Amine Amyar; Romain Modzelewski; Su Ruan

In this study, three datasets including one thousand and forty four CT images are used.The first one is a public available data set coming from [17] which includes 347 COVID-19 images and 397 non-COVID images with different kinds of pathology. The database was pre-processed and stored in png format. The dimension varies from 153 to 1853 with an average of 491 for the height, while the width varies from 124 to 383 with an average of 1485 (see Fig 3) . The second dataset coming from http://medicalsegmentation.com/covid19/ in which 100 COVID-19 CT scan with lesion ground truths are available. Three lesion labels are provided : ground glass, consolidation and plural effusion. As all legion labels are not given in all images, for the purpose of this study, we merged the three labels into one lesion label (See Fig 2) . The third dataset coming from the hospital "Henri Becquerel Center" in Rouen city of France includes 100 CT of normal patients and 98 of lung cancer. All the three image datasets were resized to have the same size of 256 x 256 and the intensity normalized between 0 and 1 prior to analysis. Table 1 summarizes how to split the datasets for training, validation and test. All procedures performed in this study were conducted according to the principles expressed in the Declaration of Helsinki. The study was approved as a retrospective study by the Henri Becquerel Center Institutional Review Board. All patient information was de-identified and anonymized prior to analysis.