AWDiff: An a trous wavelet diffusion model for lung ultrasound image synthesis

Maryam Heidari; Nantheera Anantrasirichai; Steven Walker; Rahul Bhatnagar; Alin Achim

cs.CVMar 3, 2026

AWDiff: An a trous wavelet diffusion model for lung ultrasound image synthesis

Authors: Maryam Heidari, Nantheera Anantrasirichai, Steven Walker, Rahul Bhatnagar, Alin Achim

Organizations: University of Bristol, UK · Visual Information Lab, University of Bristol, Bristol, UK · Bristol Medical School, University of Bristol, UK · Academic Respiratory Unit, Bristol Medical School, University of Bristol, Bristol, UK

Abstract

Lung ultrasound (LUS) is a safe and portable imaging modality, but the scarcity of data limits the development of machine learning methods for image interpretation and disease monitoring. Existing generative augmentation methods, such as Generative Adversarial Networks (GANs) and diffusion models, often lose subtle diagnostic cues due to resolution reduction, particularly B-lines and pleural irregularities. We propose A trous Wavelet Diffusion (AWDiff), a diffusion based augmentation framework that integrates the a trous wavelet transform to preserve fine-scale structures while avoiding destructive downsampling. In addition, semantic conditioning with BioMedCLIP, a vision language foundation model trained on large scale biomedical corpora, enforces alignment with clinically meaningful labels. On a LUS dataset, AWDiff achieved lower distortion and higher perceptual quality compared to existing methods, demonstrating both structural fidelity and clinical diversity.

Figures & tables

Figure 1: Overview of the proposed AWDiff framework. In the forward diffusion stage, the input image is decomposed by a multi-scale a trous wavelet encoder into wavelet planes (wps) that preserve fine anatomical detail. These wavelet features are then fused with BioMedCLIP embeddings during the reverse diffusion process.

Method	Steps	SIFID $\downarrow$	LPIPS $\uparrow$	NIMA $\uparrow$
SinDDM	80k	0.05 $\pm$ 0.08	0.26 $\pm$ 0.04	5.12 $\pm$ 0.5
SinDDM	120k	0.04 $\pm$ 0.01	0.27 $\pm$ 0.02	5.38 $\pm$ 0.1
SinGAN	Final (16k)	0.08 $\pm$ 0.01	0.21 $\pm$ 0.06	4.3 $\pm$ 0.1
AWDiff	80k	0.03 $\pm$ 0.5	0.36 $\pm$ 0.09	5.24 $\pm$ 0.6
AWDiff	120k	0.03 $\pm$ 0.06	0.37 $\pm$ 0.02	5.45 $\pm$ 0.7

Table 1: Comparison of SIFID, LPIPS, and NIMA for SinDDM, SinGAN, and AWDiff.

Figure 2: Comparison of AWDiff, SinDDM, and SinGAN on four training images, with three generated samples per method. AWDiff more faithfully preserves B-line structures compared to SinDDM and SinGAN.

Figure 3: LPIPS score for AWDiff, SinDDM, and SinGAN.

Figure 4: CW-SSIM comparison between à trous and DWT.

Medical Image Synthesis

Medical image synthesis creates, translates, reconstructs, and edits clinical images to fill missing data and support analysis.

See all 310 papers

Hybrid Diffusion Model for Breast Ultrasound Image Augmentation
Mar 27, 2026Farhan Fuad Abir, Sanjeda Sara Jennifer, Niloofar Yousefi +1Data Augmentation Synthetic Data Augmentation
Electrical and Computer Engineering, University of Central Florida · Department of Medicine, University of Central Florida College of Medicine · Department of Computer Science, University of Central Florida +1
Feature-Space Guided Diffusion for Realistic Ultrasound Image Synthesis
Jul 13, 2026Marina Domínguez, Nélida Mirabet-Herranz, Valery NaranjoDiffusion Model Guidance Diffusion Guidance
Instituto Universitario de Investigación en Tecnología Centrada en el Ser Humano, HUMAN-tech, Universitat Politècnica de València, Valencia, Spain · Artikode Intelligence S.L., Valencia, Spain
Controllable Lung Nodule Synthesis via Histogram-Regularized Latent Diffusion Models
May 28, 2026Arunkumar Kannan, Yanbo Zhang, Han Liu +5Synthetic Data Augmentation Latent Diffusion Models
aDigital Technology and Innovation, Siemens Healthineers, Princeton, NJ, USA · bJohns Hopkins University, Baltimore, MD, USA · cDigital Technology and Innovation, Siemens Healthineers, Erlangen, Germany +1

Mar 27, 2026J/K move · Enter open · S save

Hybrid Diffusion Model for Breast Ultrasound Image Augmentation

Farhan Fuad Abir, Sanjeda Sara Jennifer, Niloofar Yousefi, Laura J. Brattain

Electrical and Computer Engineering, University of Central Florida · Department of Medicine, University of Central Florida College of Medicine · Department of Computer Science, University of Central Florida+1

Data Augmentation Synthetic Data Augmentation Medical Image Synthesis Diffusion-Based Image Generation

Open paper AskPDF arXiv

We propose a hybrid diffusion-based augmentation framework to overcome the critical challenge of ultrasound data augmentation in breast ultrasound (BUS) datasets. Unlike conventional diffusion-based augmentations, our approach improves visual fidelity and preserves ultrasound texture by combining text-to-image generation with image-to-image (img2img) refinement, as well as fine-tuning with low-rank adaptation (LoRA) and textual inversion (TI). Our method generated realistic, class-consistent images on an open-source Kaggle breast ultrasound image dataset (BUSI). Compared to the Stable Diffusion v1.5 baseline, incorporating TI and img2img refinement reduced the Frechet Inception Distance (FID) from 45.97 to 33.29, demonstrating a substantial gain in fidelity while maintaining comparable downstream classification performance. Overall, the proposed framework effectively mitigates the low-fidelity limitations of synthetic ultrasound images and enhances the quality of augmentation for robust diagnostic modeling.

Abstract

Medical Image Synthesis

Explore similar work

Hybrid Diffusion Model for Breast Ultrasound Image Augmentation

Feature-Space Guided Diffusion for Realistic Ultrasound Image Synthesis

Controllable Lung Nodule Synthesis via Histogram-Regularized Latent Diffusion Models