<i>ai-corona</i> : Radiologist-Assistant Deep Learning Framework for COVID-19 Diagnosis in Chest CT Scans

Abstract

Abstract Generation of medical assisting tools using recent artificial intelligence advances is beneficial for the medical workers in the global fight against COVID-19 outbreak. In this article we introduce ai-corona , a radiologist-assistant deep learning framework for COVID-19 infection diagnosis using the chest CT scans. Our framework incorporates an Efficient NetB3-based feature extractor. We employed three independent dataset in this work named: CC-CCII, MDH, and MosMedData; all includes 7184 scans from 5693 subjects which contained pneumonia, common pneumonia (CP), non-pneumonia, normal and COVID-19 classes. We evaluated ai-corona on test sets from the CC-CCII set and MDH cohort and the entirety of the MosMedData cohort, for which it gained AUC score of 0.997, 0.989, and 0.954, respectively. We further compared our framework’s performance with other deep learning models developed on our employed data sets, as well as RT-PCR. Our results show that ai-corona outperforms all. Lastly, our framework’s diagnosis capabilities was evaluated as assistant to several experts. We demonstrated an increase in both speed and accuracy of expert diagnosis when incorporating ai-corona ’s assistance.

Ähnliche Arbeiten