Precise angle estimation of capsule robot in ultrasound using heatmap guided two-stage network

Area of Science:

• Medical Robotics
• Medical Imaging
• Artificial Intelligence

Background:

• Capsule robots offer non-invasive diagnosis and treatment within the gastrointestinal (GI) tract, controlled externally by magnetic fields.
• Precise angle feedback is crucial for capsule robot locomotion control, typically achieved via ultrasound imaging.
• Ultrasound-based angle estimation is challenged by factors like gastric wall tissue and internal stomach contents (air, fluid, matter).

Purpose of the Study:

• To develop an advanced method for accurate capsule robot position detection and angle estimation from ultrasound images.
• To overcome limitations of existing ultrasound-based methods in complex GI environments.

Main Methods:

• Introduction of a heatmap-guided two-stage network for capsule robot localization and angle estimation.
• Implementation of a probability distribution module for enhanced position detection.
• Utilization of skeleton extraction for precise angle calculation.

Main Results:

• The proposed network demonstrated a minimal position center error of 0.48 mm on a porcine stomach ultrasound dataset.
• Achieved a high angle estimation accuracy of 96.32% in experimental trials.
• Validated the effectiveness of the probability distribution module and skeleton extraction techniques.

Conclusions:

• The developed heatmap-guided network provides accurate angle feedback essential for capsule robot locomotion control.
• This method enhances the precision and reliability of non-invasive GI diagnosis and treatment using capsule robots.