Texture-Aware StarGAN for CT data harmonisation

2025-03-19

AI Summary

CT reconstruction kernel discrepancies induce texture inconsistencies, severely limiting the generalization capability of deep learning models. To address this, we propose the first texture-aware StarGAN framework tailored for CT data harmonization, enabling one-to-many style transfer and standardization across reconstruction kernels. Our method innovatively introduces a multi-scale texture loss that jointly models spatial- and angular-domain texture features to explicitly suppress kernel-dependent artifacts. It further integrates multi-scale LPIPS constraints, adversarial loss, and cycle-consistency loss, implemented with a ResNet-based generator and PatchGAN discriminator. Evaluated on 197 cases comprising 48,667 chest CT slices reconstructed with three distinct kernels, our approach achieves a 12.3% improvement in SSIM and a 28.7% reduction in FID. Radiologist assessments confirm statistically significant superiority over baseline StarGAN. This work establishes a novel, interpretable, and robust paradigm for CT image standardization.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Kernel Methods
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Computed Tomography (CT) plays a pivotal role in medical diagnosis; however, variability across reconstruction kernels hinders data-driven approaches, such as deep learning models, from achieving reliable and generalized performance. To this end, CT data harmonization has emerged as a promising solution to minimize such non-biological variances by standardizing data across different sources or conditions. In this context, Generative Adversarial Networks (GANs) have proved to be a powerful framework for harmonization, framing it as a style-transfer problem. However, GAN-based approaches still face limitations in capturing complex relationships within the images, which are essential for effective harmonization. In this work, we propose a novel texture-aware StarGAN for CT data harmonization, enabling one-to-many translations across different reconstruction kernels. Although the StarGAN model has been successfully applied in other domains, its potential for CT data harmonization remains unexplored. Furthermore, our approach introduces a multi-scale texture loss function that embeds texture information across different spatial and angular scales into the harmonization process, effectively addressing kernel-induced texture variations. We conducted extensive experimentation on a publicly available dataset, utilizing a total of 48667 chest CT slices from 197 patients distributed over three different reconstruction kernels, demonstrating the superiority of our method over the baseline StarGAN.

Problem

• Address variability in CT reconstruction kernels
• Enhance deep learning model generalization in medical imaging
• Improve texture-aware harmonization across different CT data sources

Innovation

• Texture-aware StarGAN for CT harmonization
• Multi-scale texture loss function integration
• One-to-many translations across reconstruction kernels