Convolutional neural network-based surgical instrument detection.

Area of Science:

• Medical Imaging
• Computer Vision
• Surgical Technology

Background:

• Minimally invasive surgery (MIS) relies on indirect visualization of surgical sites.
• Accurate instrument detection is vital for conventional and robot-assisted MIS.
• Instrument detection aids in surgical training and procedure analysis.

Purpose of the Study:

• To propose a novel framework for surgical instrument detection in MIS videos.
• To utilize a series of two three-layer convolutional neural networks (CNNs) for this task.
• To enhance visualization and safety in minimally invasive procedures.

Main Methods:

• A two-step approach using two CNNs was developed.
• The first CNN detects instrument shaft edge points from image patches.
• The second CNN locates the instrument tip following the initial edge detection.

Main Results:

• The framework was validated on the EndoVisSub and a standard dataset.
• Detection accuracies of 91.2% and 75% were achieved, respectively.
• The proposed method demonstrated superior performance compared to existing approaches.

Conclusions:

• The two-step CNN detection method offers high accuracy for surgical instruments.
• This approach improves upon current methods for instrument detection in MIS.
• The framework has potential applications in surgical training and real-time guidance.