Predefined and data driven CT densitometric features predict critical illness and hospital length of stay in COVID-19 patients

Abstract

The aim of this study was to compare whole lung CT density histograms to predict critical illness outcome and hospital length of stay in a cohort of 80 COVID-19 patients. CT chest images on segmented lungs were retrospectively analyzed. Functional Principal Component Analysis (FPCA) was used to find the main modes of variations on CT density histograms. CT density features, the CT severity score, the COVID-GRAM score and the patient clinical data were assessed for predicting the patient outcome using logistic regression models and survival analysis. ROC analysis predictors of critically ill status: 87.5th percentile CT density (Q875)—AUC 0.88 95% CI (0.79 0.94), F1-CT—AUC 0.87 (0.77 0.93) Standard Deviation (SD-CT)—AUC 0.86 (0.73, 0.93). Multivariate models combining CT-density predictors and Neutrophil–Lymphocyte Ratio showed the highest accuracy. SD-CT, Q875 and F1 score were significant predictors of hospital length of stay (LOS) while controlling for hospital death using competing risks models. Moreover, two multivariate Fine-Gray regression models combining the clinical variables: age, NLR, Contrast CT factor with either Q875 or F1 CT-density predictors revealed significant effects for the prediction of LOS incidence in presence of a competing risk (death) and acceptable predictive performances (Bootstrapped C-index 0.74 [0.70 0.78]).

Quantile Correlation‐Based Sufficient Variable Screening by Controlling False Discovery Rate

“…In parallel, the second method, known as the percentile density measurement (PD) technique, brings forth a novel dimension to our analysis. It gracefully dissects the distribution curve of lung density attenuation and provides a window into the percentile values (often the 5th and 95th percentiles) Shalmon et al [29] With this foundation in place, this section focus turns to the meticulous analysis of a lung CT image dataset, downloaded from the platform Kaggle. This invaluable dataset, meticulously segmented to isolate the lungs accurately, becomes the canvas on which we shall apply our proposed method.…”

“…It gracefully dissects the distribution curve of lung density attenuation and provides a window into the percentile values (often the 5th and 95th percentiles) Shalmon et al. [ 29 ]…”

Quantile-Adaptive Sufficient Variable Screening by Controlling False Discovery

“…The other is the percentile density measurement (PD) technique. Analyze the attenuation distribution curve of lung density, give a percentile (commonly 5% and 95%), calculate the area below the percentile density curve, and evaluate the symptoms of emphysema [ 28 ]. In this section, we shall apply our proposed method to analyze the lung CT image dataset downloaded from Kaggle , which can segment lungs accurately.…”

Assessment of pulmonary fibrosis severity at autopsy in patients after COVID-19: comparison with quantitative CT scan data in the acute phase of the disease

“…The development of quantitative processing of CT images allowed us to build models of pneumonia severity stratification [41,42], prediction of length of hospital stay and probability of death [43].…”