Six different datasets were curated with clinical information from two different countries, namely Brazil and Italy. Different haematology parameters were trained to develop machine learning models, based on the datasets. Deep neural network failed on these datasets, as there were a large number of null entries. Hence, machine learning model have been developed based on XGBoost classifier. The performances of the models based on the above six datasets were compared with the performance of reported models.
Four hematology parameters identified, namely, leukocyte, eosinophil, monocyte and platelet count, can distinguish COVID-19 infection from five other similar respiratory viruses, namely, influenza A (H1N1), influenza B, coronavirus N63, rhinovirus and respiratory syntactical virus. Statistical significance tests validated the differences between comorbidities.
Birla Institute of Technology
Assistant Professor
Acces sibility Controls