PredRA: Fast Medical Image Translation by Deterministic Component Extraction and Controlled Stochastic Refinement

2026-09-25

AI Summary

This study addresses the inherent tension in medical image translation between deterministic predictions, which often lose fine details, and generative models, which introduce undesirable variations. To resolve this, we propose a novel framework that integrates determinism with controlled stochasticity. Specifically, the method first extracts deterministic image components to ensure baseline fidelity, then formulates residual generation as an optimization problem. A global control mechanism suppresses uncertainty accumulation, while a flow matching sampler combined with a deep residual network enables efficient refinement. Experimental results demonstrate that the proposed approach significantly improves structural fidelity while preserving perceptual quality. Furthermore, it reduces parameter counts by 1.4× to 11.9× and decreases sampling steps by 31.25×, achieving fast, high-fidelity medical image translation with substantially lower computational overhead.

Abstract

Strongly paired medical image translation contains a substantial component that can be predicted directly from the source. The current reality is that pure deterministic prediction can smooth away fine detail, while generative models can recover detail but may also introduce unnecessary or potentially harmful variation. We propose PredRA, a fast framework that uses the deterministic prediction as a stable reference and further extracts additional deterministic components from the residual during generative refinement, thereby improving fidelity while maintaining perceptual quality. The goal is simple: we view the entire residual-based generative process as an optimization problem and derive a practical solution that recovers useful residual detail through controlled refinement while keeping the prediction close to the paired target. Mechanism studies further show that useful residual information follows structured patterns, but its usefulness is difficult to estimate reliably at the voxel level. We therefore globally control how much of the residual refinement is added to the deterministic prediction, thereby reducing the accumulation of unnecessary uncertainty. PredRA therefore combines deterministic component extraction with controlled stochastic refinement for fast, fidelity-preserving, and perceptually strong medical image translation. We validate the approach across multiple real-world datasets, showing that controlled refinement improves fidelity to the paired target compared with full residual refinement while retaining much of the perceptual benefit of generative modeling. PredRA achieves competitive or superior performance to substantially larger state-of-the-art models with 1.4-11.9x fewer total parameters and 3.1-26.2x fewer trainable parameters, while its 32-step flow sampler requires 31.25x fewer sampling steps than the matched 1000-step DDPM.

Problem

• Medical Image Translation
• Deterministic Prediction
• Generative Refinement
• Image Fidelity
• Computational Efficiency

Innovation

• Medical Image Translation
• Deterministic Component Extraction
• Controlled Stochastic Refinement
• Flow Matching
• Parameter Efficiency