Improved brain tumor segmentation by utilizing tumor growth model in longitudinal brain MRI

Area of Science:

• Medical Imaging
• Computational Biology
• Biomedical Engineering

Background:

• Accurate brain tumor segmentation is crucial for diagnosis and treatment planning.
• Texture and intensity features in MRI are commonly used but have limitations.
• Integrating dynamic growth patterns could enhance segmentation accuracy.

Purpose of the Study:

• To develop and evaluate a novel method for brain tumor segmentation.
• To fuse cell density patterns from computational tumor growth modeling with MRI features.
• To assess the improvement in segmentation using the proposed combined approach.

Main Methods:

• Tumor growth was modeled by solving the reaction-diffusion equation using the Lattice-Boltzmann method (LBM).
• Cell density distributions were generated as novel features.
• These features were fused with fractal, multifractal Brownian motion (mBm), and intensity features from MRI.
• The method was evaluated on longitudinal MRI scans from the BRATS 2015 dataset.

Main Results:

• The proposed method demonstrated significant improvement in complete tumor segmentation.
• Validation was performed using ground truth on MRI scans from five patients.
• ANOVA analysis confirmed the statistical significance of the improvements in longitudinal MR images.

Conclusions:

• Fusing cell density patterns from tumor growth modeling with traditional MRI features enhances brain tumor segmentation.
• The Lattice-Boltzmann method provides valuable cell density information for segmentation.
• This novel approach offers a promising direction for improving automated tumor segmentation in clinical practice.