SALIENT: Frequency-Aware Paired Diffusion for Controllable Long-Tail CT Detection

2026-02-26

AI Summary

This work addresses the challenge of diminished detection accuracy for rare lesions in whole-body CT scans, which arises from extreme class imbalance and extremely low volumetric prevalence. To tackle this, the authors propose a mask-guided frequency-domain diffusion framework that performs structured diffusion in the wavelet domain, decoupling low-frequency intensity from high-frequency structural information. They introduce, for the first time, a frequency-aware objective function to disentangle lesion and background characteristics. The approach integrates a 3D VAE for lesion mask generation, a wavelet-domain diffusion model, a frequency-aware loss, and a semi-supervised teacher model that produces slice-level pseudo-labels, enabling controllable augmentation and joint optimization. Experiments demonstrate significant improvements in synthetic quality (MS-SSIM: 0.63→0.83; FID: 118.4→46.5) and substantial gains in downstream detection performance under low prevalence, with the optimal synthesis ratio adapting dynamically to the amount of available labeled data.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Detection of rare lesions in whole-body CT is fundamentally limited by extreme class imbalance and low target-to-volume ratios, producing precision collapse despite high AUROC. Synthetic augmentation with diffusion models offers promise, yet pixel-space diffusion is computationally expensive, and existing mask-conditioned approaches lack controllable attribute-level regulation and paired supervision for accountable training. We introduce SALIENT, a mask-conditioned wavelet-domain diffusion framework that synthesizes paired lesion-masking volumes for controllable CT augmentation under long-tail regimes. Instead of denoising in pixel space, SALIENT performs structured diffusion over discrete wavelet coefficients, explicitly separating low-frequency brightness from high-frequency structural detail. Learnable frequency-aware objectives disentangle target and background attributes (structure, contrast, edge fidelity), enabling interpretable and stable optimization. A 3D VAE generates diverse volumetric lesion masks, and a semi-supervised teacher produces paired slice-level pseudo-labels for downstream mask-guided detection. SALIENT improves generative realism, as reflected by higher MS-SSIM (0.63 to 0.83) and lower FID (118.4 to 46.5). In a separate downstream evaluation, SALIENT-augmented training improves long-tail detection performance, yielding disproportionate AUPRC gains across low prevalences and target-to-volume ratios. Optimal synthetic ratios shift from 2x to 4x as labeled seed size decreases, indicating a seed-dependent augmentation regime under low-label conditions. SALIENT demonstrates that frequency-aware diffusion enables controllable, computationally efficient precision rescue in long-tail CT detection.

Problem

• long-tail CT detection
• class imbalance
• rare lesion detection
• precision collapse
• synthetic data augmentation

Innovation

• frequency-aware diffusion
• wavelet-domain synthesis
• paired lesion-masking
• long-tail CT detection
• controllable augmentation