Prediction of BRAF mutation status of craniopharyngioma using magnetic resonance imaging features

Abstract

OBJECTIVE Treatment with a BRAF mutation inhibitor might shrink otherwise refractory craniopharyngiomas and is a promising preoperative treatment to facilitate tumor resection. The aim of this study was to investigate the noninvasive diagnosis of BRAF-mutated craniopharyngiomas based on MRI characteristics. METHODS Fifty-two patients with pathologically diagnosed craniopharyngioma were included in this study. Polymerase chain reaction was performed on tumor tissue specimens to detect BRAF and CTNNB1 mutations. MRI manifestations-including tumor location, size, shape, and composition; signal intensity of cysts; enhancement pattern; pituitary stalk morphology; and encasement of the internal carotid artery-were analyzed by 2 neuroradiologists blinded to patient identity and clinical characteristics, including BRAF mutation status. Results were compared between the BRAF-mutated and wild-type (WT) groups. Characteristics that were significantly more prevalent (p < 0.05) in the BRAF-mutated craniopharyngiomas were defined as diagnostic features. The minimum number of diagnostic features needed to make a diagnosis was determined by analyzing the receiver operating characteristic (ROC) curve. RESULTS Eight of the 52 patients had BRAF-mutated craniopharyngiomas, and the remaining 44 had BRAF WT tumors. The clinical characteristics did not differ significantly between the 2 groups. Interobserver agreement for MRI data analysis was relatively reliable, with values of Cohen κ ranging from 0.65 to 0.97 (p < 0.001). A comparison of findings in the 2 patient groups showed that BRAF-mutated craniopharyngiomas tended to be suprasellar (p < 0.001), spherical (p = 0.005), predominantly solid (p = 0.003), and homogeneously enhancing (p < 0.001), and that patients with these tumors tended to have a thickened pituitary stalk (p = 0.014). When at least 3 of these 5 features were present, a tumor might be identified as BRAF mutated with a sensitivity of 1.00 and a specificity of 0.91. The area under the ROC curve for the sum of all 5 diagnostic criteria was 0.989 (p < 0.001). CONCLUSIONS The BRAF mutation status of craniopharyngiomas might be predicted using certain MRI features with relatively high sensitivity and specificity, thus offering potential guidance for the preoperative administration of BRAF mutation inhibitors.

Noninvasive molecular diagnosis of craniopharyngioma with MRI-based radiomics approach

“…Deep learning-based radiomics (DLR) was developed on a dataset of 151 patients with low-grade glioma with multiple modalities of magnetic resonance (MR) images for predicting the mutation status of IDH1, and achieved ACC of 0.91 and AUC of 0.96 [16]. Yue et al proposed a novel diagnostic criterion of BRAF mutation and wild type in craniopharyngiomas with SENS of 1.00 and SPEC of 0.91 [18]. In our study, high-throughput features applied into our radiomics model are extracted quantitatively and noninvasively on medical images following the well-established standard to describe the structural heterogeneity of tumors.…”

“…Immunostaining for BRAF V600E (Spring Biosciences, USA, VE-1, 1:40) and β-Catenin (BD Biosciences, USA, clone 14, 1:800) was performed by Ventana NexES Staining System in a proportion of patients. BRAF V600E on exon 15 and CTNNB1 mutation on exon 3 were detected and analyzed by sanger sequencing as previously reported [18].…”

“…The location of craniopharyngioma was considered highly correlating with its genotype [18]. Instead of visually determining whether the tumor is completely suprasellar as previous study [18], a more quantitative location evaluation method was used in this study.…”

“…The location of craniopharyngioma was considered highly correlating with its genotype [18]. Instead of visually determining whether the tumor is completely suprasellar as previous study [18], a more quantitative location evaluation method was used in this study. To reduce the anatomical variability among individual brain, Statistical Parametric Mapping package (SPM) was utilized to register the MR images into the normalized Montreal Neurological Institute(MNI) space [19].…”

Targeted treatment of papillary craniopharyngiomas harboring BRAF V600E mutations

“…In addition, the remarkable responses in patients PCP1 to PCP5 with recurrent disease and in PCP6 with neoadjuvant treatment after biopsy signal the potential for using targeted therapy before any surgery. Algorithms for discriminating PCPs from adamantinomatous craniopharyngiomas have been proposed on the basis of retrospective analyses of MRI data . Such approaches will require further refinement and testing in prospective studies.…”