PURPOSE: To evaluate the potential role of an Artficial Intelligence (AI) software, Contextflow (contextflow Gmbh, Wien, Austria), to support radiologists in distinguishing between COVID-19 and non-COVID-19 patients.
METHODS: For this single-center retrospective study, we will enroll a cohort of patients admitted at the Emergency Department of Sant'Andrea Hospital in Rome from March 4, 2020 to March 31, 2020, who underwent unenhanced chest-CT. All suspected COVID-19 patients enrolled were tested with two nasopharyngeal and oropharyngeal swabs, analyzed with RT-PCR (Charitè, Berlin, Germany). All chest-CT scans will be anonymized and uploaded from the hospital's Picture Archiving and Communication Systems (PACS) to the dedicated server of Contextflow. Uploaded CT images will be automatically analyzed by AI system in order to detect eighteen COVID-19 pneumonia features (ground glass opacities, parenchymal consolidations, etc), providing quantitative volumetric measurements, expressed in percentage. All data elaborated by the software will be analized by two radiologists in consensus.
Dedicated MedCalc software (MedCalc Software,version 15, Ostend, Belgium) and Statistical Package for Social Sciences (SPSS 23.0, IBM, Chicago, USA) will be used for statistical analysis and results will be expressed as mean±standard deviation. The results will be statistically significant when p
EXPECTED RESULTS: We expect that Contextflow will identify accurately the most frequent COVID-19 pneumonia features and we hope that the obtained results will make possible to differentiate between positive and negative COVID-19 patients with quantitative values.
EXPECTED CONCLUSIONS: AI software, Contextflow, will support radiologists in chest CT quantification to distinguish between COVID-19 and non-COVID-19 patients, with an higher sensibility and specificity.
Several differences between chest CT of COVID-19 and non-COVID-19 patients are emerging in literature, but these data are not yet well established. AI could support in distinguishing between these two groups of patients, in a quantitative way, increasing diagnostic accuracy and thus have a pivotal clinical role in patients management.