Toward Image Data-Driven Predictive Modeling for Guiding Thermal Ablative Therapy.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Modeling

Background:

• Accurate prospective modeling of microwave ablation (MWA) is crucial for procedural planning and navigation.
• Patient-specific tissue properties, vital for accurate modeling, are often unavailable and vary due to factors like perfusion and disease state.
• Existing modeling frameworks require enhancement to account for these tissue property variations.

Purpose of the Study:

• To establish an inverse modeling approach for reconstructing tissue properties.
• To create a predictive model of tissue properties as a function of fat content.
• To evaluate the accuracy and improve the utility of predictive procedural modeling for MWA.

Main Methods:

• An inverse modeling approach was used to reconstruct tissue properties by fitting model predictions to observed ablation zone extents in phantoms.
• A model correlating tissue properties with fat content was developed and evaluated using leave-one-out cross-validation.
• Predictions were validated using phantoms with co-recorded temperature data.

Main Results:

• The model-based approach achieved thermal profiles closely matching experimental measurements (average RMSE of 4.8°C).
• Model-predicted ablation zones demonstrated significant overlap with observed ablations (93.4 ± 2.2% in validation, 86.6 ± 5.3% in cross-validation).
• An average 17.3% improvement in predicted ablation zone overlap was observed compared to models using component volume fractions.

Conclusions:

• The study demonstrates accurate ablation estimates using image-driven determination of tissue properties.
• The developed inverse modeling approach successfully links physical modeling parameters with quantitative medical imaging.
• This work serves as a proof-of-concept for enhancing predictive procedural modeling in MWA.