Prior-Guided Residual Diffusion: Calibrated and Efficient Medical Image Segmentation
This addresses uncertainty in medical image segmentation for healthcare applications, representing an incremental improvement over existing diffusion methods.
The paper tackled ambiguity in medical image segmentation by introducing Prior-Guided Residual Diffusion (PGRD), a diffusion-based framework that captures full conditional distributions, resulting in higher Dice scores and lower NLL/ECE values compared to baselines while requiring fewer sampling steps.
Ambiguity in medical image segmentation calls for models that capture full conditional distributions rather than a single point estimate. We present Prior-Guided Residual Diffusion (PGRD), a diffusion-based framework that learns voxel-wise distributions while maintaining strong calibration and practical sampling efficiency. PGRD embeds discrete labels as one-hot targets in a continuous space to align segmentation with diffusion modeling. A coarse prior predictor provides step-wise guidance; the diffusion network then learns the residual to the prior, accelerating convergence and improving calibration. A deep diffusion supervision scheme further stabilizes training by supervising intermediate time steps. Evaluated on representative MRI and CT datasets, PGRD achieves higher Dice scores and lower NLL/ECE values than Bayesian, ensemble, Probabilistic U-Net, and vanilla diffusion baselines, while requiring fewer sampling steps to reach strong performance.