MRI transformer deep learning and radiomics for predicting IDH wild type TERT promoter mutant gliomas.

in NPJ precision oncology by Wenju Niu, Junyu Yan, Min Hao, Yibo Zhang, Tianshi Li, Chen Liu, Qijian Li, Zihao Liu, Yincheng Su, Bo Peng, Yan Tan, Xiaochun Wang, Lei Wang, Hui Zhang, Guoqiang Yang

TLDR

  • A new computer model can accurately predict patient outcomes for a particular type of brain cancer, using MRI scans and ensemble learning techniques.
  • The model outperforms other models and can help doctors make more informed treatment decisions.
  • The study's findings have the potential to improve patient care and outcomes in neuro-oncology.

Abstract

This study aims to predict IDH wt with TERTp-mut gliomas using multiparametric MRI sequences through a novel fusion model, while matching model classification metrics with patient risk stratification aids in crafting personalized diagnostic and prognosis evaluations.Preoperative T1CE and T2FLAIR sequences from 1185 glioma patients were analyzed. A MultiChannel_2.5D_DL model and a 2D DL model, both based on the cross-scale attention vision transformer (CrossFormer) neural network, along with a Radiomics model, were developed. These were integrated via ensemble learning into a stacking model. The MultiChannel_2.5D_DL model outperformed the 2D_DL and Radiomics models, with AUCs of 0.806-0.870. The stacking model achieved the highest AUC (0.855-0.904) across validation sets. Patients were stratified into high-risk and low-risk groups based on stacking model scores, with significant survival differences observed via Kaplan-Meier analysis and log-rank tests. The stacking model effectively identifies IDH wt TERTp-mutant gliomas and stratifies patient risk, aiding personalized prognosis.

Overview

  • The study aims to predict IDH wt with TERTp-mut gliomas using multiparametric MRI sequences through a novel fusion model.
  • The study analyzed preoperative T1CE and T2FLAIR sequences from 1185 glioma patients.
  • The study developed and integrated three different models: MultiChannel_2.5D DL, 2D DL, and Radiomics, using ensemble learning and stacking to achieve the highest AUC.

Comparative Analysis & Findings

  • The MultiChannel_2.5D_DL model outperformed the 2D_DL and Radiomics models, with AUCs of 0.806-0.870.
  • The stacking model achieved the highest AUC (0.855-0.904) across validation sets.
  • Patients were stratified into high-risk and low-risk groups based on stacking model scores, with significant survival differences observed via Kaplan-Meier analysis and log-rank tests.

Implications and Future Directions

  • The study's findings can aid personalized prognosis and diagnosis for IDH wt TERTp-mutant gliomas.
  • Future studies can explore using the developed models for other types of gliomas and incorporating additional imaging sequences and biomarkers.
  • The study highlights the importance of developing accurate and comprehensive diagnostic models to improve patient outcomes and treatment decisions.
Read Full Article