Automatic contour tracking in ultrasound images

Area of Science:

• Medical imaging
• Biomedical engineering
• Computer vision

Background:

• Ultrasound imaging of the human tongue presents challenges due to noise and high-contrast edges.
• Accurate tracking of tongue surfaces is crucial for speech production research and diagnostics.

Purpose of the Study:

• To present EdgeTrak, a novel automatic contour tracking system for human tongue ultrasound sequences.
• To develop an improved active contour model capable of handling noise and open contours in ultrasound images.

Main Methods:

• Development of a novel active contour model incorporating edge gradient and local intensity information.
• Utilizing contour orientation to discard irrelevant edges.
• Employing dynamic programming for optimization.
• Application to ultrasound image sequences from a head and transducer support system (HATS).

Main Results:

• The proposed active contour model effectively tracks tongue surfaces in ultrasound images, outperforming classical models.
• EdgeTrak demonstrates robustness and accuracy in identifying partial tongue surfaces, even with open contours.
• Quantitative comparison validates the system's performance against tracking by speech scientists.

Conclusions:

• The EdgeTrak system provides a reliable and accurate method for automatic contour tracking of human tongue ultrasound images.
• The novel active contour model enhances tracking capabilities by integrating local image information and orientation.
• This advancement has significant implications for speech science research and clinical applications.