Statistical analysis
Quantitative variables are reported as mean ± standard deviation and
qualitative data as number and percentage. The survival curves were
established by the Kaplan-Meier method. Multivariate analyses were
performed to identify independent predictors of OS and EFS. To take into
account the correlation between the estimates of each texture parameter
from the different filter values as well as the small number of events
compared with the number of included covariates, multivariate L1 (least
absolute shrinkage and selection operator—Lasso) penalized Cox
regression models logistic regression models were built in order to
select clinical and texture parameters(7). The regularization parameter
was determined by using fivefold cross-validation. The Lasso method
allows variable selection by shrinking down to zero coefficient weights
for variables non-related to outcome. Variables with non-zero
coefficients were selected as potential predictors of outcome and
integrated into multivariable Cox regression analyses in order to
estimate associated hazard ratios (HR) and their 95% confidence
intervals (CI 95%).
For each texture parameter associated with outcome in multivariate
analysis, a receiver operating characteristic (ROC) curve was
constructed to identify the most relevant threshold.
A p value < 0.05 was considered statistically significant.
Analyses were performed using SAS version 9.4 (SAS Inc, Cary, NC, USA)
and R version 3.6 (R Foundation for Statistical Computing, Vienna,
Austria).