Computer-Assisted Differentiation between Colon-Mesocolon and Retroperitoneum Using Hyperspectral Imaging (HSI)...

Area of Science:

• Surgical Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Complete mesocolic excision (CME) offers superior oncological outcomes but carries a higher risk of vascular injury.
• Accurate identification of surgical planes is crucial for safe CME.
• Hyperspectral imaging (HSI) provides non-invasive, quantitative tissue characterization.

Purpose of the Study:

• To evaluate the accuracy of HSI combined with deep learning (DL) in differentiating colonic/mesenteric tissue from retroperitoneal tissue.
• To assess the potential of HSI-DL for intraoperative guidance during CME.
• To develop an automated, objective method for surgical plane identification.

Main Methods:

• An animal study utilizing 20 pig models.
• Intraoperative hyperspectral imaging of the sigmoid colon, mesentery, and retroperitoneum.
• Training a convolutional neural network (CNN) on HSI data for tissue classification.
• Validation using a leave-one-out cross-validation approach.

Main Results:

• The CNN achieved an overall recognition sensitivity of 79.0 ± 21.0% for retroperitoneal tissue (to be preserved).
• The CNN achieved an overall recognition sensitivity of 86.0 ± 16.0% for colon and mesenteric tissue (to be resected).
• HSI and CNN demonstrated accurate, non-invasive tissue differentiation.

Conclusions:

• Automated classification using HSI and CNNs shows promise for objective, non-invasive differentiation of surgical planes.
• This technology can potentially enhance the safety and precision of complete mesocolic excision.
• HSI-DL offers a novel approach to intraoperative anatomical structure visualization.