Diffusion-weighted MRI precisely predicts telomerase reverse transcriptase promoter mutation status in World Health Organization grade IV gliomas using a residual convolutional neural network

Abstract

Abstract Objectives Telomerase reverse transcriptase promoter (pTERT) mutation status plays a key role in making decisions and predicting prognoses for patients with World Health Organization (WHO) grade IV glioma. This study was conducted to assess the value of diffusion-weighted imaging (DWI) for predicting pTERT mutation status in WHO grade IV glioma. Methods MRI data and molecular information were obtained for 266 patients with WHO grade IV glioma at the hospital and divided into training and validation sets. The ratio of training to validation set was approximately 10:3. We trained the same residual convolutional neural network (ResNet) for each MR modality, including structural MRIs (T1-weighted, T2-weighted, and contrast-enhanced T1-weighted) and DWI*, to compare the predictive capacities between DWI and conventional structural MRI. We also explored the effects of different regions of interest on pTERT mutation status prediction outcomes. Results Structural MRI modalities poorly predicted the pTERT mutation status (accuracy = 51%-54%; area under the curve [AUC]=0.545-0.571), whereas DWI combined with its apparent diffusive coefficient maps yielded the best predictive performance (accuracy = 85.2%, AUC = 0.934). Including the radiological and clinical characteristics did not further improve the performance for predicting pTERT mutation status. The entire tumour volume yielded the best prediction performance. Conclusions DWI technology shows promising potential for predicting pTERT mutations in WHO grade IV glioma and should be included in the MRI protocol for WHO grade IV glioma in clinical practice. Advances in knowledge This is the first large-scale model study to validate the predictive value of DWI for pTERT in WHO grade IV glioma.