Endoscopic stereo reconstruction: a comparative study

Area of Science:

• Medical image analysis
• Robotic surgery
• Computer vision

Background:

• Minimally-invasive surgery (MIS) relies on endoscopic imagery.
• Computer-vision algorithms are increasingly vital for surgical decision support.
• Stereo reconstruction methods are well-studied for general environments but less so for endoscopic data.

Purpose of the Study:

• To extensively compare state-of-the-art stereo reconstruction methods for endoscopic images.
• To analyze the advantages and disadvantages of these methods within a surgical context.
• To provide a comprehensive evaluation for advancing computer vision in MIS.

Main Methods:

• Comparison of several state-of-the-art stereo reconstruction algorithms.
• Testing on both simulated and real endoscopic images.
• Utilizing data from a daVinci stereo endoscope with controlled in-lab and phantom models.

Main Results:

• Detailed analysis of the strengths and weaknesses of each stereo reconstruction method.
• Evaluation across key stages of the stereo reconstruction pipeline.
• Experimental validation on diverse endoscopic datasets.

Conclusions:

• Stereo reconstruction is essential for enhancing computer vision in endoscopic surgery.
• This comparative study provides critical insights into method selection for MIS.
• Findings support the development of more robust surgical decision support systems.