Model Adaptation for Image Reconstruction using Generalized Stein's Unbiased Risk Estimator
This incremental improvement addresses model adaptation issues for medical imaging researchers and practitioners in multi-site or varied contrast settings.
The paper tackled the problem of model mismatches in deep learning image reconstruction when acquisition schemes differ from training, by introducing a Generalized Stein's Unbiased Risk Estimator loss metric to adapt networks to measured data and reduce misfit impact, resulting in improved reconstruction quality compared to mean-square error methods.
Deep learning image reconstruction algorithms often suffer from model mismatches when the acquisition scheme differs significantly from the forward model used during training. We introduce a Generalized Stein's Unbiased Risk Estimate (GSURE) loss metric to adapt the network to the measured k-space data and minimize model misfit impact. Unlike current methods that rely on the mean square error in kspace, the proposed metric accounts for noise in the measurements. This makes the approach less vulnerable to overfitting, thus offering improved reconstruction quality compared to schemes that rely on mean-square error. This approach may be useful to rapidly adapt pre-trained models to new acquisition settings (e.g., multi-site) and different contrasts than training data