Lung CT Scan Analysis of SARS-CoV2 Induced Lung Injury

BACKGROUND: In February, the first case of SARS-CoV2 positive patient was recorded in Lombardy (Italy), a virus capable of causing a severe form of acute respiratory failure called Coronavirus Disease 2019 (COVID-19). Qualitative assessments of lung morphology have been identified to describe macroscopic characteristics of this infection upon admission and during the hospitalization of patients. At the moment, there are no studies that have exhaustively described the parenchymal lung damage induced by SARS-CoV2 by quantitative analysis. The hypothesis of this study is that specific morphological and quantitative alterations of the lung parenchyma assessed by means of CT scan in patients suffering from severe respiratory insufficiency induced by SARS-CoV2 may have an impact on the severity of the degree of alteration of the respiratory exchanges (oxygenation and clearance of the CO2) and have an impact on patient outcome. The presence of characteristic lung morphological patterns assessed by CT scan could allow the recognition of specific patient clusters who can benefit from intensive treatment differently, making a significant contribution to stratifying the severity of patients and their risk of mortality. This is an exploratory clinical descriptive study of lung CT images in a completely new patient population who are nucleic acid amplification test confirmed SARS-CoV2 positive. SAMPLE SIZE (n. patients): The study will collect all patients with the inclusion criteria; a total of 500 patients are expected to be collected. About 80 patients will be enrolled for each local experimental center. The following patient data will be analyzed: * blood gas analytical data assigned to the CT scan, checks performed upon entering the hospital, at the time of performing the CT scan, admission to intensive care and 7 days after entry * patient characteristics such as age, gender and body mass index (BMI) * comorbidity * presence of organ dysfunction with the Sequential Organ Failure Assessment (SOFA) * laboratory data relating to hospital admission and symptoms prior to hospitalization. * ventilator and hemodynamic parameters upon entering the hospital, at the time of carrying out the CT scan, upon admission to intensive care and 7 days after entry. The machine learning approach of lung CT scan analysis will aim at evaluating: 1. Quantitative and qualitative lung alterations; 2. The stratification of such morphological characteristics in specific morphological lung clusters identified by the means of artificial intelligence using deep learning algorithms. ETHICAL ASPECTS: The lung CT scan images will be collected and anonymized. Images will be subsequently sent by University of Milano-Bicocca Institutional google drive account to the University of Pennsylvania, Department of Anesthesiology and Critical Care and the Department of Radiology in a deidentified format for advanced quantitative analysis taking advantage of artificial intelligence using deep learning algorithms. The data will be collected in a pseudo-anonymous way through paper Case Report Form (CRF) and analyzed by the scientific coordinator of the project. Given the retrospective nature of the study and in the presence of technical difficult in obtaining an informed consent of patients in this period of pandemic emergency, informed consent will be waived. STATISTICAL ANALYSIS: Continuous data will be expressed as mean ± standard deviation or median and interquartile range, according to data distribution that will be evaluated by the Shapiro-Wilk test. Categorical variables will be expressed as proportions (frequency). The deep learning segmentation algorithm will segment the lung parenchyma from the entire CT lung. Lung volume, lung weight and opacity intensity distribution analysis will be applied. Second, clustering analysis to stratify the patients will be performed. Both an intensity and a spatial clustering algorithm will be tested. Third, a model will be trained to predict the injury progression using the images and all other patient data. Statistical significance will be considered in the presence of a p\<0.05 (two-tailed).