Robust tip localization under continuous spatial and temporal constraints during 2D ultrasound-guided needle puncture

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Ultrasound-guided (US-guided) needle puncture is crucial for minimally invasive procedures.
• Automated needle tip localization in US images aids clinicians in precise needle guidance.
• Interference from echoes poses a significant challenge to accurate localization.

Purpose of the Study:

• To develop a robust automated needle tip localization method for real-time US-guided procedures.
• To enhance clinician confidence and precision during needle punctures.
• To overcome challenges posed by echo interferences in US images.

Main Methods:

• A novel method utilizing continuous spatial and temporal constraints in real-time US frame streams.
• Employs deep learning-based detectors for motion-enhanced US images and raw grayscale B-mode images.
• Predicts tip locations using estimated tip velocity for frames lacking constraints.

Main Results:

• Achieved 97.2% recall and 91.9% precision on porcine organ data, with 0.88 ± 0.70 mm RMSE.
• On unseen human thyroid data, obtained 86.5% recall, 84.3% precision, and 0.92 ± 0.78 mm RMSE.
• Demonstrated efficient performance with a running speed of 14.5 frames per second on CPU.

Conclusions:

• The proposed method offers a reliable solution for automated needle tip localization in US-guided procedures.
• The technique is robust to interferences commonly encountered in ultrasound imaging.
• Fast running speed ensures practicability for real-time US applications.