Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
This work evaluates Dormant Awakening (DA)—a parameter-efficient adaptation technique that tunes a tiny subset of zero-initialized weights on a single labeled calibration slice—augmented with output-scaling mechanisms to safely adapt MRI reconstruction models across domain shifts. Across 14 U-Net and Vision Transformer backbones evaluated on a fastMRI-to-M4Raw transfer, fitting between 1,004 and 10,560 parameters produces mean PSNR improvements of 0.430 dB and 0.509 dB, respectively. By dynamically capping the output correction based on calibration sizes, the method successfully bounds test-time RMSE degradation relative to the source checkpoint without requiring ground-truth references, outperforming unconstrained LoRA and sparse adaptation.
Tuning as few as 1,000 dormant weights on a single labeled slice delivers a 0.5 dB PSNR boost in MRI reconstruction, paired with dynamic output capping that provably bounds drift under domain shift without ground-truth references.
Dormant Awakening (DA) specializes a magnetic resonance imaging (MRI) reconstruction model by fitting a small set of zero weights to one labeled slice. Across fourteen U-shaped convolutional network (U-Net) and vision transformer (ViT) sources, fitting 1,004 or 10,560 weights gives mean peak signal-to-noise ratio (PSNR) gains of .430 and .509 dB. We analyze how scaling the fitted output correction changes evaluation PSNR, then use the calibration correction size to cap changes on new images. The cap bounds changes in root mean squared error (RMSE) relative to the source without an evaluation reference. At 10\% adaptation budget, we compare fixed halving and dynamic capping for DA, unrestricted sparse adaptation and low-rank adaptation (LoRA). The controls increase both measured calibration-evaluation correlations across all six tested architecture-adapter settings. Experiments concern one constructed fastMRI-to-M4Raw shift with fixed source checkpoints and participant panels.