Machine learning-based prognostic subgrouping of glioblastoma: A multicenter study

Area of Science:

• Oncology
• Medical Imaging
• Machine Learning

Background:

• Glioblastoma (GBM) is a highly aggressive brain cancer with significant patient heterogeneity.
• This heterogeneity complicates patient management, treatment planning, and clinical trial stratification.

Purpose of the Study:

• To develop a reproducible, personalized prognostication and clinical subgrouping system for glioblastoma.
• To leverage machine learning on routine clinical and imaging data for improved patient stratification.

Main Methods:

• Developed a machine learning model using routine clinical data, MRI, and molecular measures from 2838 diverse patients across 22 institutions.
• Stratified patients into favorable, intermediate, and poor prognostic subgroups (I, II, III) using Kaplan-Meier analysis and Cox proportional models.

Main Results:

• The ML model successfully stratified patients into distinct prognostic subgroups with significant hazard ratios.
• Imaging features provided unique prognostic value, supporting a generalizable prognostic classification system.

Conclusions:

• The ML model is reproducible and accessible online, using routine imaging data.
• This platform facilitates personalized patient management and clinical trial stratification for glioblastoma.