Abstract
Intraoperative 2D-3D spine registration is a critical step in Image-Guided Spinal Surgery, requiring accurate alignment of preoperative CT volumes with intraoperative C-Arm X-ray images under stringent real-time constraints. In this paper, we propose SpineSAC, a Hierarchical Multi-agent Soft Actor-Critic Reinforcement Learning frame work for autonomous six degree of freedom vertebral pose-based 2D 3D registration. SpineSAC integrates a Principal Component Analysis (PCA)-based Canonical Coordinate System (CCS) to establish a common anatomical reference frame across different CT and C-Arm coordinate frames and a novel Rotational Feature Encoder (RFE) that encodes Antero-Posterior and Lateral DRR and X-ray rotational misalignment into a compact 16-dimensional latent observation for the RL agent. SpineSAC achieves a Total Registration Error (TRE) of 1.45 ± 0.5mm on phantoms and 2.31mm on cadaver data, with a 9× inference speedup over the computational approach, demonstrating its suitability for real time intraoperative deployment. The Code is available at: SpineSAC.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/COLAS_020.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=pOobSrh7La
BibTex
@InProceedings{K.Dar_SpineSAC_MICCAISAT2026,
author = { K. N., Darshan AND L., Sriram AND P., Sarath AND Maik, Vivek AND Lakshmanan, Manojkumar AND Sivaprakasam, Mohanasankar},
title = { { SpineSAC: Hierarchical Multi-Agent Soft Actor-Critic for Patient-Agnostic 2D–3D Spine Registration in Image-Guided Spine Surgery } },
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}
}