Robust adaptive extended Kalman filtering for real time MR-thermometry guided HIFU interventions
Sébastien Roujol1, Baudouin Denis de Senneville, Silke Hey
1FRE 3313 CNRS/University Victor Segalen Bordeaux, Bordeaux, France. sebastien.roujol@imf.u-bordeaux2.fr
IEEE Transactions on Medical Imaging
|October 15, 2011
Summary
A new adaptive filter improves real-time magnetic resonance (MR) thermometry precision for high-intensity focused ultrasound (HIFU) tumor ablation. This advanced filtering enhances thermal dose measurement accuracy, even with temperature artifacts, outperforming traditional methods.
Area of Science:
- Medical Physics
- Biomedical Engineering
- Radiology
Background:
- Real-time magnetic resonance (MR) thermometry is crucial for guiding high-intensity focused ultrasound (HIFU) tumor ablations.
- Fast MR imaging required for thermometry often suffers from low signal-to-noise ratios (SNRs), limiting precision.
- Existing temporal filtering methods can negatively impact accuracy and introduce latency.
Purpose of the Study:
- To develop and evaluate a novel adaptive filter for enhancing MR thermometry precision during HIFU ablation.
- To improve the accuracy and robustness of thermal dose measurements in the presence of physiological motion and artifacts.
- To overcome the limitations of low SNRs in fast MR acquisition schemes for real-time thermometry.
Main Methods:
- An adaptive extended Kalman filter was developed, incorporating a heat transfer model for acoustic heating in biological tissues.
- An outlier rejection mechanism was added to handle sparse, artifacted temperature data points.
- The novel filter was compared against an efficient matched Finite Impulse Response (FIR) filter using simulated data, phantom experiments, and in vivo porcine kidney studies.
Main Results:
- The adaptive filter demonstrated substantial accuracy improvements: a factor of 3 during heat-up and 15 during cool-down in worst-case simulations.
- The filter proved robust during HIFU experiments, unaffected by strong temperature artifacts.
- In contrast, the FIR filter exhibited a high measurement variation of 70% under similar artifact conditions.
Conclusions:
- The proposed adaptive extended Kalman filter significantly enhances the precision and robustness of real-time MR thermometry for HIFU applications.
- This novel approach effectively mitigates the impact of low SNRs and temperature artifacts, improving thermal dose measurement accuracy.
- The filter's ability to adapt and reject outliers makes it a promising tool for sophisticated HIFU control algorithms and clinical feasibility.
