Estimating Glioblastoma Biophysical Growth Parameters Using Deep Learning Regression.

Area of Science:

• Computational oncology
• Artificial intelligence in medicine
• Neuro-oncology imaging

Background:

• Glioblastoma (GBM) is a highly aggressive and heterogeneous adult brain tumor.
• Biophysical modeling aids clinical decisions but is computationally intensive, taking hours.
• Accelerating GBM growth parameter estimation is crucial for clinical application.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for rapid estimation of GBM biophysical growth parameters.
• To significantly reduce the computation time for GBM growth modeling from hours to seconds.
• To enable faster clinical translation of biophysical models for glioblastoma.

Main Methods:

• A deep learning (DL)-based logistic regression model was developed to estimate GBM growth parameters.
• Three key parameters were estimated: white matter diffusion coefficient (Dw), mass-effect parameter (Mp), and tumor growth time (T).
• The DL model was trained and validated using pre-operative multi-parametric MRI (mpMRI) scans from 135 TCGA-GBM subjects.

Main Results:

• The DL model achieved high accuracy in estimating biophysical growth parameters.
• Average Pearson correlation coefficients were 0.85 for Dw, 0.90 for Mp, and 0.94 for T.
• The DL approach reduced computation time from hours to seconds, a significant acceleration.

Conclusions:

• Deep learning enables rapid and accurate estimation of glioblastoma biophysical growth parameters.
• This accelerated approach facilitates clinical translation of GBM growth models.
• The study opens avenues for leveraging radiomic descriptors through faster parameter reconstruction.