Radiogenomics-Based Risk Prediction of Glioblastoma Multiforme with Clinical Relevance.

Area of Science:

• Neuro-oncology
• Radiomics and Radiogenomics
• Biostatistics and Predictive Modeling

Background:

• Glioblastoma multiforme (GBM) is an aggressive brain tumor with challenging treatment monitoring.
• Temozolomide (TMZ)-based radiochemotherapy can cause pseudo-progression (PsP), mimicking true tumor progression on MRI.
• Accurate prognostication is crucial for effective GBM patient management.

Purpose of the Study:

• To develop a prognostic model for evaluating glioblastoma multiforme tumor progression potential post-standard therapies.
• To differentiate true tumor progression (TTP) from pseudo-progression (PsP) using imaging and genomic data.
• To establish a robust predictor of 2-year survival (2YS) in GBM patients.

Main Methods:

• Applied dictionary learning to extract imaging features from GBM patients (Wake dataset).
• Conducted radiogenomics analysis to identify genes associated with imaging features.
• Constructed a 2YS logistic regression model using significantly associated genes and validated it on The Cancer Genome Atlas Program (TCGA) datasets.

Main Results:

• The developed 2YS model effectively classified GBM patients into low- and high-survival risk groups.
• 2YS scores were significantly associated with overall patient survival in both training and independent testing TCGA datasets.
• Clinical factors like age, gender, Karnofsky performance status (KPS), and normal cell ratio showed weak or no correlation with survival.

Conclusions:

• The 2YS model demonstrates effectiveness and robustness in predicting clinical outcomes for GBM patients.
• This radiogenomic approach offers a reliable method for prognostication, aiding in clinical decision-making.
• The model's performance surpasses that of traditional clinical factors in predicting GBM patient survival.