Artificial Intelligence-Based Total Mesorectal Excision Plane Navigation in Laparoscopic Colorectal Surgery.

Area of Science:

• Surgical Oncology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Total mesorectal excision (TME) is the gold standard for rectal cancer surgery, crucial for minimizing local recurrence.
• Identifying the correct dissection plane in TME can be challenging for surgeons.

Purpose of the Study:

• To develop and evaluate the feasibility of a deep learning-based image-guided navigation system for TME.
• To utilize areolar tissue as a visual landmark for accurate dissection plane identification.

Main Methods:

• A single-center feasibility study involving 32 patients undergoing laparoscopic colorectal resection.
• Development of a deep learning model for semantic segmentation of areolar tissue in the TME plane using intraoperative video data.
• Evaluation of segmentation accuracy using the Dice coefficient.

Main Results:

• A deep learning model was successfully developed to locate and highlight areolar tissue in the TME plane.
• The system demonstrated high accuracy in identifying the TME plane, potentially aiding surgical dissection.
• Limited training data (600 images) necessitates further data acquisition for improved accuracy.

Conclusions:

• A novel image-guided navigation system for TME, based on areolar tissue segmentation, was successfully developed.
• This system shows promise in assisting surgeons to accurately identify the TME plane, potentially improving surgical outcomes.
• Future work should focus on increasing the training dataset size to enhance model performance.