Real-time vascular anatomical image navigation for laparoscopic surgery: experimental study.

Area of Science:

• Surgical Innovation
• Artificial Intelligence in Medicine
• Vascular Anatomy

Background:

• Inferior mesenteric artery (IMA) recognition is vital in colorectal cancer surgery for hemorrhage control and lymph node dissection.
• Accurate IMA identification aids in defining surgical margins and improving patient outcomes.
• Laparoscopic colorectal surgery necessitates advanced tools for precise anatomical identification.

Purpose of the Study:

• To develop a deep learning model for anatomical recognition of the IMA.
• To evaluate the accuracy and real-time performance of the developed IMA recognition model.
• To assess the feasibility of using deep learning for real-time vascular navigation in laparoscopic colorectal surgery.

Main Methods:

• Utilized a multi-institutional surgical video database (LapSig300).
• Extracted intraoperative videos from 60 patients undergoing laparoscopic sigmoid colon resection or high anterior resection.
• Employed deep learning-based semantic segmentation and evaluated using Dice Similarity Coefficient (DSC) and frames per second (FPS).

Main Results:

• Achieved a mean DSC of 0.798 (±0.0161 SD) and a maximum DSC of 0.816 in fivefold cross-validation.
• The deep learning model demonstrated real-time performance exceeding 12 FPS.
• The model showed relatively high accuracy in recognizing the IMA in intraoperative images.

Conclusions:

• This is the first study to evaluate real-time vascular anatomical navigation using deep learning for IMA recognition in laparoscopic colorectal surgery.
• The developed deep learning model shows feasibility and potential for image navigation systems in laparoscopic procedures.
• While safety and usefulness require clinical verification, the model's accuracy in IMA recognition is promising for surgical navigation.