Predicting intratumoral fluid pressure and liposome accumulation using physics informed deep learning.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Oncology

Background:

• Liposome-based anticancer agents utilize the enhanced permeability and retention (EPR) effect for tumor targeting.
• Clinical success of nano-therapeutics is limited by factors like high interstitial fluid pressure (IFP), which impedes drug delivery and predicts treatment efficacy.
• Accurate measurement of liposome distribution and IFP within tumors is vital for effective treatment planning.

Purpose of the Study:

• To develop a predictive model for voxel-by-voxel intratumoral liposome accumulation and IFP.
• To utilize pre- and post-administration imaging data for model input.
• To apply a novel physics-informed machine learning approach combining machine learning and partial differential equations.

Main Methods:

• Development of a physics-informed machine learning model.
• Integration of machine learning with partial differential equations.
• Validation using both mouse data and synthetically generated tumors.

Main Results:

• The developed model accurately predicts spatial liposome accumulation and IFP within individual tumors.
• The approach requires minimal input information for accurate predictions.
• Demonstrated efficacy on both experimental mouse data and simulated tumor models.

Conclusions:

• The physics-informed machine learning model offers a powerful tool for predicting intratumoral liposome accumulation and IFP.
• This predictive capability is crucial for optimizing cancer treatment strategies.
• The approach has significant potential for forecasting tumor progression and guiding personalized treatment design.