Preference-guided CT projection correction for surface-based feather morphometrics

Traditional imaging pipelines often optimize for human-readable images and generic quality metrics, but algorithm-facing morphometric workflows require preservation of the information used by downstream measurement. Acquisitions that appear overexposed or visually nonstandard may therefore be appropriate when they expose subtle structures to an analysis algorithm. We developed a task-aligned projection-correction framework for reducing morphology-relevant CT motion artifacts without relying on human visual interpretability as the final quality standard. Using CT-based feather morphometry as a motivating example, we treated a previously published feather-growth computer-vision workflow as the downstream task to be protected during correction. A residual denoising autoencoder defined a bounded projection-correction space, and an expert human-preference ranker scored candidate residual corrections for preservation of task-relevant morphology. In a five-scene synthetic proof-of-concept benchmark comprising four development seeds and one fully held-out seed, rank-weighted evolution-strategy updates optimized latent corrections under projection-preservation, latent-bound, and angular-smoothing constraints. The development-inclusive five-scene summary showed 67.9% lower reconstruction RMSE and 72.0% lower density-map RMSE than motion-corrupted data, with the fully held-out seed showing the same direction and similar magnitude of improvement. Preference-guided correction also outperformed a feed-forward residual-autoencoder baseline (reconstruction RMSE 0.00260 ± 0.00013; density-map RMSE 0.0500 ± 0.0021), whereas TomoPy joint reprojection alignment changed the principal morphology-error endpoints by approximately 1% or less under the local non-rigid feather-motion benchmark. Clean-data controls produced only small perturbations when no motion correction was needed. In two illustrative real-data case studies, the same framework reduced visible feather ghosting and improved downstream crest-density mapping and agreement with manual reference masks when inserted upstream of the published morphometric workflow. These results support task-aligned projection correction as a route toward CT imaging workflows optimized for downstream biological measurement rather than generic visual image quality alone. The study provides proof-of-concept evidence that expert preference signals can guide bounded projection-domain correction for surface-based biological morphometrics.