Patient-Specific Real-Time Segmentation in Trackerless Brain Ultrasound

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Intraoperative ultrasound (iUS) offers potential for improved brain surgery outcomes.
• Interpreting iUS is challenging for neurosurgeons.
• Accurate brain tumor segmentation is crucial for surgical planning and execution.

Purpose of the Study:

• To develop the first patient-specific framework for trackerless intraoperative ultrasound brain tumor segmentation.
• To adapt iUS interpretation to neurosurgical objectives in real-time.
• To improve the accuracy and reliability of brain tumor segmentation during surgery.

Main Methods:

• A patient-specific real-time network was designed for brain tumor segmentation in trackerless iUS.
• Synthetic ultrasound data was generated by simulating virtual iUS sweep acquisitions from pre-operative MRI.
• The network was trained using this synthetic data to disambiguate iUS and adapt to surgical goals.

Main Results:

• The proposed framework demonstrated effectiveness in segmenting brain tumors using real intraoperative ultrasound data.
• The approach successfully adapted to surgeons' definitions of surgical targets.
• Patient-specific models outperformed non-patient-specific models, expert neurosurgeons, and high-end tracking systems.

Conclusions:

• The developed patient-specific framework significantly enhances brain tumor segmentation in trackerless iUS.
• This technology has the potential to improve surgical navigation and outcomes in neurosurgery.
• The framework offers a adaptable and high-performing solution for real-time surgical guidance.