Enhancing handheld ultrasound images with style-driven data synthesis
Style-Driven Data Synthesis and Degradation-Aware Enhancement for Ultrasound Image Restoration
Computer Vision and Pattern Recognition
Summary
Low-cost handheld ultrasound machines create images that are often blurry or unclear compared to professional hospital machines, which can cause mistakes in diagnosis. The authors address this by creating a two-step process: first, they generate pairs of blurry and clear images that match exactly in layout, even though real scans don’t. Then, they train a model to improve the blurry images based on these pairs. Their approach showed better results in making handheld ultrasound images clearer and more accurate.
What this means in practice
- •For medical imaging teams: Improve the diagnostic quality of handheld ultrasound images by training enhancement models with synthetic pixel-aligned datasets derived from hospital-quality scans.
- •For software developers for medical devices: Incorporate degradation-aware image enhancement techniques in portable ultrasound device software to deliver clearer images without needing exact paired training data.$Commercial implications: Enables sale of smarter ultrasound enhancement features for handheld devices improving clinical usability and image quality.
Authors
Yu-Kai Wang, Chun-Xin Tan, Manh-Hung Nguyen, Ching-Chun Huang
Abstract
Low-cost handheld ultrasound devices can be widely deployed compared to professional hospital ultrasound machines. However, their images suffer from compound degradation that can mislead clinical judgment. Motivated by this observation, mapping handheld low-quality (LQ) to hospital high-quality (HQ) images has been considered a valuable research question. Conventionally, the mapping requires pixel-aligned LQ-HQ pairs. This requirement is unsatisfactory in practical scenarios because real scans at different times are never pixel-aligned. This paper addresses the challenge with a two-stage framework. The first stage generates pixel-aligned LQ-HQ datasets, and the second stage trains an enhancement model that improves LQ images. The first stage trains a cycle-consistent style-transfer model on unaligned real LQ-HQ pairs to learn a HQ-to-LQ model. Then, the model transforms real HQ images into pixel-aligned LQ images. Based on the dataset generated by the first stage, the second stage uses the Dual Degradation-Guided (DDG) Low-Rank Adaptation (LoRA) method to fine-tune an LQ-to-HQ model based on aligned pairs. In this stage, the model is based on the well known PiSA-SR framework but inserts a degradation-conditioned correction matrix. Experimental results on the USenhance2023 dataset show that the FID metric is improved by 16.7% over the strongest baseline while other metrics indicate that our enhanced outputs are well aligned with the real HQ distribution. The source code of our method is available at https://github.com/Jason0411202/DDG_LoRA.