Thermal Ablation Therapy Control with Tissue Necrosis-driven Temperature Feedback Enabled by Neural State Space Model...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Biology

Background:

• Thermal ablation is a key minimally invasive treatment, but monitoring its progression and temperature is challenging.
• Current methods for visualizing ablation and sensing temperature have limitations, including sensitivity to motion and environmental factors, risking patient overheating.
• Relying solely on necrosis monitoring can lead to dangerous tissue overheating.

Purpose of the Study:

• To develop a necrosis progression-based temperature estimation technique for thermal ablation therapy.
• To create an added safety feature to prevent tissue overheating during minimally invasive treatments.
• To propose a model-based method that does not require additional sensing hardware.

Main Methods:

• Utilized a Neural State Space model to approximate the nonlinear partial differential equations governing ablation therapy.
• Designed an Extended Kalman Filter based on the Neural State Space model for temperature estimation.
• Conducted simulation studies to evaluate the robustness and accuracy of the estimation module under various noise conditions.

Main Results:

• The developed estimation module demonstrated robust tissue temperature estimation, even with different types of noise.
• The maximum observed estimation error before terminating ablation was approximately 1 °C.
• The proposed safety feature for preventing overheating during thermal ablation was successfully demonstrated.

Conclusions:

• The necrosis progression-based temperature estimation technique offers a precise and safe approach to thermal ablation therapy.
• This model-based method can function independently or complement existing thermometers, enhancing robustness.
• The underlying architecture is generalizable to other medical/biological procedures with complex physics or lacking reliable theoretical models.