Contrast-Enhanced MRI Texture Parameters as Potential Prognostic Factors for Primary Central Nervous System Lymphoma Patients Receiving High-Dose Methotrexate-Based Chemotherapy

Abstract EN

Introduction.

The purpose of this study was to evaluate the prognostic value of texture features on contrast-enhanced magnetic resonance imaging (MRI) for patients with primary central nervous system lymphoma (PCNSL).

Methods.

In this retrospective study, fifty-two patients diagnosed with PCNSL were enrolled from October 2010 to March 2017.

The texture feature of tumor tissue on the histogram-based matrix (histo-) and the grey-level co-occurrence matrix (GLCM) was retrieved by contrast-enhanced T1-weighted imaging before any antitumor treatment.

Receiver operating characteristic curve analyses were performed to obtain their optimal cutoff values, based on which we dichotomized patients into subgroups.

The Kaplan–Meier analyses were conducted to compare overall survival (OS) of subgroups, and multivariate Cox regression analyses were used to determine if they could be taken as independent prognostic factors.

Results.

Ten texture features were extracted from the MR image, including Energy, Entropy, Kurtosis, Skewness on the histogram-based matrix, and Correlation, Contrast, Dissimilarity, Energy, Entropy, and Homogeneity on the grey-level co-occurrence matrix.

Three of them (GLCM-Contrast, GLCM-Dissimilarity, and GLCM-Homogeneity) are shown to be significant in relation to overall survival (OS).

The multivariate Cox regression analyses suggest that GLCM-Homogeneity could be taken as independent predictors.

Conclusions.

The texture features of contrast-enhanced magnetic resonance imaging (MRI) could potentially serve as prognostic biomarkers for PCNSL patients.