Real-Time Reconstruction of HIFU Focal Temperature Field Based on Deep Learning

Area of Science:

• Medical Physics
• Biomedical Engineering
• Oncology

Background:

• High-intensity focused ultrasound (HIFU) offers noninvasive treatment for diseased tissues via thermal and cavitation effects.
• Accurate, real-time temperature monitoring during HIFU therapy is crucial for enhancing efficacy and minimizing damage to healthy tissues.
• Current methods for temperature field acquisition during HIFU treatment require improvement for clinical application.

Purpose of the Study:

• To develop and validate a novel multimodal teacher-student model for real-time temperature field reconstruction during HIFU therapy.
• To integrate diagnostic, therapeutic, and temperature measurement functionalities into a single HIFU system.
• To provide precise predictive schemes for personalized HIFU cancer treatment planning.

Main Methods:

• Designed and assembled an integrated HIFU system for collecting ultrasound echo signals and temperature variations.
• Introduced a multimodal teacher-student model utilizing shared self-expressive coefficients and deep canonical correlation analysis.
• Employed knowledge distillation strategies to transfer information from a teacher model to a student model.

Main Results:

• Successfully achieved real-time 2D temperature field reconstruction in the HIFU focal region.
• Demonstrated a maximum temperature error of less than 2.5 °C across phantoms, in vitro, and in vivo studies.
• Validated the correlation between ultrasound echo signals and temperature variations.

Conclusions:

• The developed method enables effective real-time monitoring of the HIFU temperature field distribution.
• Provides scientifically precise predictive schemes for HIFU therapy, supporting personalized treatment dose planning.
• Offers efficient guidance for noninvasive, nonionizing cancer treatment, laying a foundation for future clinical applications.