Abstract
Reliable real-time semantic segmentation of laparoscopic video is critical for intraoperative decision support. Nonetheless, frame-wise predictions are prone to noisy temporal flickering, while tissue deformation and scene motion require responsiveness to local changes. Although exponential moving average (EMA) filtering serves as a model-agnostic and lightweight post-processing strategy for mitigating such flickering, its fixed coefficient imposes a trade-off between temporal stability and motion responsiveness. To address this limitation, we propose a local motion-adaptive temporal smoothing method that replaces the fixed EMA coefficient with a spatially varying coefficient map derived from block-wise aggregation of sparse optical flow without requiring training or architectural modification. In this method, block-wise local motion estimated from sparse optical flow is converted to adaptive coefficients through a smoothstep mapping, and the resulting coefficients are applied online to segmentation probability maps. Experiments on the CholecSeg8k dataset show that the proposed method improves temporal stability in low-motion regions while preserving frame-wise accuracy in high-motion regions. Qualitative results demonstrate reduced flickering in stable regions without visible lag in dynamic regions. Runtime analysis shows a 90th-percentile latency of 9.41 ms, well below the 30-fps frame interval. These results support the potential of the proposed method for stabilizing real-time surgical video segmentation in intraoperative decision-support systems.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SafeSurg_017.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/SafeSurg_017_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=46vmgVZnpQ
BibTex
@InProceedings{YamMin_Local_MICCAISAT2026,
author = { Yamashita, Minato AND Ito, Masaaki},
title = { { Local Motion-Adaptive Temporal Smoothing for Real-Time Laparoscopic Video Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS pending},
month = {pending},
page = {pending}
}