42P Radiation induced pneumonitis in the era of the COVID-19 pandemic: Artificial intelligence for differential diagnosis

Abstract

Background: In 2019, the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) was identified in Wuhan, China. The most serious clinical entity associated with SARS-CoV-2 is a severe interstitial pneumonia that can lead to acute respiratory distress. Radiation pneumonitis (RP) is lung radiation toxicity. RP and SARS-CoV-2 interstitial pneumonia show overlapping clinical features. The aim of this study is to test the performance of a deep learning algorithm in discriminating RP from COVID-19 pneumonia. Methods: Seventy patients were analysed, 34 affected by COVID-19 pneumonia and 36 by RP. The CT images were analyzed by InferReadTM CT Lung (COVID-19) ®, an Artificial Intelligence algorithm based on a novel deep convolutional neural network structure. In a previous publication the cut-off value of the estimated risk probability of COVID-19 was set at levels higher than 30% ("COVID19 High Risk"), as the percentage of COVID-19 confirmed patients above this cut-off value was higher than 95%. Values of estimated risk probability below 30% were classified as "COVID19 Low Risk". Statistical analysis included Mann Whitney U test (significance threshold at p < 0.05) and ROC curve with fitting performed by using the maximum likelihood fit of a binormal model. Results:The algorithm classified as "COVID19 Low Risk" 66.7% of patients presenting RP. All RP classified as "COVID19 High Risk" were ≥G3. The algorithm showed good accuracy in the detection of RP against COVID-19 pneumonia (sensitivity = 97.0%, specificity = 2%, AUC = 0.72). This accuracy increased when cut-off of 30% was applied (sensitivity 76%, specificity 63%, AUC = 0.84. The total lung volume (p = 0.001), the left lower lobe ( p < 0.001) and the right lower lobe (p < 0.001) involvement resulted increased in COVID-19 group compared to RP. In patients pretreated with radiotherapy and presenting diffuse pneumonitis classified by AI as COVID19 High Risk a combination of dosimetric factors may help to identify RP (PPV increased from 60% to 99.8%). Conclusions: Deep-learning algorithm can help to discriminate RP from COVID-19 pneumonia, classifying most RP as Low-risk COVID19. In patients classified as high risk, treated with radiation therapy also dosimetric factors should be taken into account.