Improved bone X-ray images by using decoder information in AI models
RFS-UNet: Decoder-Conditioned High-Resolution Skip Recalibration for Bone-Selective DRR Synthesis
Computer Vision and Pattern Recognition
Summary
Creating detailed synthetic bone X-ray images relies on combining information from different parts of an AI model. The researchers found that including information from the part of the model that builds the output (called the decoder) helps reuse image details better than just using the part that analyzes the input (called the encoder). Their approach, RFS-UNet, improves image quality slightly without making the model much bigger or slower. This means the method can produce clearer bone images from scans more reliably.
digitally reconstructed radiograph (DRR)encoderdecoderskip connectionsresidual channel recalibrationPSNR (peak signal-to-noise ratio)MAE (mean absolute error)convolutional neural network (Conv2d)CT scanpaired projection synthesis
Authors
Xiaoyang Li, Yixuan Liu, Yuan Chai
Abstract
Bone-selective digitally reconstructed radiograph (DRR) synthesis depends on high-resolution encoder detail, yet static skips cannot condition reuse on the evolving decoder representation. We ask whether decoder state adds useful information beyond encoder-only self-recalibration for high-resolution skip reuse. RFS-UNet uses pooled encoder and aligned decoder statistics for bounded residual channel recalibration at the 512^2 and 256^2 skips, leaving the backbone unchanged. In the matched seed-2026 comparison isolating decoder conditioning, RFS raises validation PSNR by 0.254 dB over Self-RFS. Across three seeds, locked-test PSNR rises from 33.225+/-0.048 to 33.537+/-0.128 dB; RFS lowers MAE in 179/200 held-out CT cases and reduces mean MAE by 3.91%. It adds 0.117% parameters and 1.169% counted Conv2d operations. These results support decoder state as a useful conditioning signal for high-resolution feature reuse in controlled paired projection synthesis.