IMAGE-DRIVEN BIOPHYSICAL TUMOR GROWTH MODEL CALIBRATION.

Area of Science:

• Computational biology
• Medical imaging
• Biophysics

Background:

• Biophysical tumor growth models often require multiple time-point data for accurate calibration.
• Glioblastoma multiforme (GBM) necessitates precise modeling for effective treatment planning.
• Magnetic Resonance Imaging (MRI) provides essential multiparametric data for tumor characterization.

Purpose of the Study:

• To develop a novel formulation for calibrating a biophysical tumor growth model from a single-time snapshot, multiparametric MRI scan of a glioblastoma patient.
• To create a two-snapshot scenario using an atlas as a substitute for a pretumor scan, enabling parameter extraction from limited patient data.
• To enhance the accuracy and reliability of patient-specific tumor growth parameter estimation.

Main Methods:

• Utilized a PDE-constrained optimization framework and a modified Picard-iteration-type solution strategy.
• Combined image-registration and parameter estimation algorithms, simulating tumor growth in the patient brain domain.
• Incorporated an L1 sparsity constraint on the initial tumor condition and solved sub-problems using a reduced space, inexact Gauss-Newton-Krylov/quasi-Newton method.

Main Results:

• The novel scheme demonstrated more accurate and reliable reconstruction of tumor parameters compared to previous methods.
• Simulating tumor growth directly in the patient brain domain yielded more meaningful patient-specific results.
• The joint inversion scheme effectively integrated registration, parameter estimation, and sparsity constraints.

Conclusions:

• The proposed method offers a robust approach for calibrating biophysical tumor growth models from single-time MRI scans.
• This technique improves the accuracy of patient-specific glioblastoma modeling, potentially aiding in treatment strategy development.
• The integration of atlas-based data and advanced computational methods provides a reliable framework for analyzing limited clinical data.