A learning-based, region of interest-tracking algorithm for catheter detection in echocardiography

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Surgery

Background:

• Echocardiography (echo) is increasingly used for catheter guidance in surgical procedures.
• Accurate visualization of the catheter tip in echo images remains challenging, even with specialized catheters.

Purpose of the Study:

• To develop and validate an automated method for detecting acoustically active catheter tips in cardiac echo images.
• To improve the accuracy and robustness of catheter tip localization during echocardiographic guidance.

Main Methods:

• A convolutional neural network (CNN) model was trained to identify the left ventricle region of interest (ROI).
• A color intensity difference technique was applied within the ROI for catheter detection.
• Catheter tracking methods were implemented and tested using manually identified initial positions.

Main Results:

• The CNN and color intensity difference method achieved high detection rates (94% long-axis, 57% short-axis) without manual input.
• Tracking methods successfully distinguished the catheter tip from surrounding structures in subsequent frames.
• The combined approach resulted in automatic, robust, and fast catheter detection.

Conclusions:

• The developed dual-technique method offers the first automatic and reliable solution for catheter detection in echocardiography.
• This advancement has the potential to enhance the safety and efficiency of catheter-guided surgical procedures.
• Further integration of these techniques can significantly improve real-time echocardiographic guidance.