Determination of Robust Regional CT Radiomics Features for COVID-19

BackgroundThe lung CT images of COVID-19 patients can be characterized by three different regions - Ground Glass Opacity (GGO), consolidation and pleural effusion. GCOs have been shown to precede consolidations. Quantitative characterization of these regions using radiomics can facilitate accurate diagnosis, disease progression and response to treatment. However, according to the knowledge of the author, regional CT radiomics analysis of COVID-19 patients has not been carried out. This study aims to address these by determining the radiomics features that can characterize each of the regions separately and can distinguish the regions from each other. Methods44 radiomics features were generated with four quantization levels for 23 CT slice of 17 patients. Two approaches were the implemented to determine the features that can differentiate between lung regions - 1) Z-score and correlation heatmaps and 2) one way ANOVA for finding statistically significantly difference (p<0.05) between the regions. Radiomics features that show agreement for all cases (Z-score, correlation and statistical significant test) were selected as suitable features. The features were then tested on 52 CT images. Results10 radiomics features were found to be the most suitable among 44 features. When applied on the test images, they can differentiate between GCO, consolidation and pleural effusion successfully and the difference provided by these 10 features between three lung regions are statistically significant. ConclusionThe ten robust radiomics features can be useful in extracting quantitative data from CT lung images to characterize the disease in the patient, which in turn can help in more accurate diagnosis, staging the severity of the disease and allow the clinician to plan for more successful personalized treatment for COVID-19 patients. They can also be used for monitoring the progression of COVID-19 and response to therapy for clinical trials.