SpikeReg: Energy-Efficient 3D Deformable Medical Image Registration with Spiking Neural Networks

2026-05-24

AI Summary

This work addresses the high computational cost of 3D deformable medical image registration, which hinders deployment in resource-constrained settings. The authors propose SpikeReg, the first spiking neural network (SNN)-based method for 3D brain MRI registration, built upon a spiking U-Net architecture. SpikeReg is initialized via layer-wise weight transfer and activation percentile threshold calibration from an ANN teacher model, then fine-tuned using a surrogate gradient objective that combines local cross-correlation, diffusion regularization, and spike-rate sparsity. Evaluated on the OASIS dataset, SpikeReg achieves a Dice coefficient of 0.7474 ± 0.032—comparable to its ANN counterpart—while operating at an average spike rate of only 12.8%, yielding a 55.5× reduction in arithmetic energy consumption. This study demonstrates, for the first time, the feasibility of SNNs in high-dimensional medical image registration with simultaneous accuracy and energy efficiency.

Abstract

Deformable medical image registration aligns anatomical structures across images but remains computationally dense at 3D resolution. Spiking neural networks (SNNs) offer sparse event-driven computation, yet have not been systematically studied for deformable medical image registration. We introduce SpikeReg, a spiking U-Net for 3D brain MRI registration. SpikeReg is initialized from an analog ANN registration teacher, converted by layer-wise weight transfer and activation-percentile threshold calibration, and fine-tuned with a surrogate-gradient objective combining local cross-correlation, diffusion regularization, and spike-rate sparsity. On the OASIS Learn2Reg validation split ($19$ image pairs), SpikeReg reaches Dice $0.7474 \pm 0.032$, with no significant paired Dice difference from the ANN teacher ($0.7480 \pm 0.037$, $p = 0.67$), at a $12.8\%$ mean spike rate and a $55.5\times$ projected arithmetic-energy reduction under an event-sparse SynOps/MAC proxy relative to the dense-ANN baseline. We additionally report two negative findings: displacement distillation from the ANN teacher hurts performance, and ANN teachers trained with a label-Dice loss fail to transfer through rate-code conversion. Together these results show that dense geometric prediction can be performed under sparse event-driven computation, opening a path toward neuromorphic medical image registration.

Problem

• deformable medical image registration
• energy efficiency
• spiking neural networks
• 3D MRI registration
• neuromorphic computing

Innovation

• Spiking Neural Networks
• 3D Medical Image Registration
• Energy-Efficient Computing
• Surrogate Gradient Learning
• Neuromorphic Vision