High Efficiency Video Coding (HEVC)-Based Surgical Telementoring System Using Shallow Convolutional Neural Network.

Area of Science:

• Biomedical Engineering
• Computer Science

Background:

• Surgical telementoring systems face bandwidth limitations for real-time remote surgical guidance.
• High Efficiency Video Coding (HEVC/H.265) offers compression but involves a quality-bandwidth trade-off.
• Deep learning for surgical region extraction is computationally intensive, unsuitable for real-time encoding.

Purpose of the Study:

• To develop an efficient surgical telementoring system addressing bandwidth constraints.
• To propose a hybrid lossless-lossy video compression approach prioritizing surgical regions.
• To introduce a shallow convolutional neural network (S-CNN) for efficient surgical region extraction.

Main Methods:

• A shallow convolutional neural network (S-CNN) encoder-only architecture for surgical incision region extraction.
• Hybrid lossless-lossy video compression coding the surgical incision region in high quality and background in low quality.
• Comparison of S-CNN segmentation performance against SegNet and evaluation of HEVC encoding with the proposed scheme.

Main Results:

• S-CNN achieved 97% pixel accuracy and 79% mean intersection over union for segmentation.
• The proposed HEVC encoding scheme reduced bitrate by an average of 88.8% compared to full-frame HEVC.
• An average signal-to-noise ratio gain of 11.5 dB was observed in the surgical region.

Conclusions:

• The proposed S-CNN based segmentation and hybrid coding scheme is effective for surgical telementoring.
• The system maintains high visual quality in critical surgical areas while achieving significant bitrate savings.
• This approach offers an efficient solution for surgical telementoring in low-bandwidth network environments.