MRI features for prediction of the intravenous chemotherapy effect in patients with retinoblastoma

Abstract

To investigate the value of orbital magnetic resonance imaging (MRI) features in predicting the efficacy of intravenous chemotherapy (IVC) for patients with retinoblastoma (RB). The pretreatment clinical and MRI data of 100 eyes from 80 RB patients who underwent IVC were collected retrospectively. There were 59 eyes in the effective group and 41 eyes in the ineffective group, and the baseline data of the two groups were compared statistically. Three radiologists reviewed and evaluated each lesion independently based on 25 MRI features. The predictive values of the MRI features for IVC efficacy were assessed by multi-factor logistic regression analysis, and their odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated. Receiver operating characteristic curves (ROCs) with the area under the curve (AUC) were used to determine the predictive abilities. A predictive model was constructed by integrating all independent predictors visualised by the nomogram. There were no statistically significant differences in sex or age between the effective and ineffective groups. The results of multivariate regression analysis showed that laterality, margin, and anterior eye segment enhancement were identified as independent factors that could predict IVC efficacy. The predictive model combining these three features was constructed, and it had an AUC of 0.732 (95% CI: 0.633, 0.831, p<0.01), a sensitivity of 71.2%, and a specificity of 70.7%. The data demonstrate that the orbital MRI features can be used to predict IVC efficiency before RB patients are treated.