Lung-DDPM+: Efficient Thoracic CT Image Synthesis using Diffusion Probabilistic Model

2025-08-12

AI Summary

Existing generative models for lung cancer diagnosis suffer from low computational efficiency and poor anatomical fidelity. To address this, we propose a semantic layout-guided denoising diffusion probabilistic model (DDPM) tailored for high-fidelity synthetic CT image generation of pulmonary nodules. Methodologically, we innovatively integrate semantic layout priors of pulmonary nodules with an anatomy-aware DPM-solver sampling acceleration strategy specifically designed for thoracic imaging. Evaluated on the LIDC-IDRI dataset, our approach reduces FLOPs by 8×, decreases GPU memory consumption by 6.8×, and accelerates sampling by 14×, while preserving high anatomical consistency. Downstream pulmonary nodule segmentation achieves state-of-the-art performance. To the best of our knowledge, this is the first work to jointly introduce semantic layout guidance and organ-specific sampling acceleration into medical image synthesis—significantly enhancing clinical applicability through improved efficiency, fidelity, and task-oriented utility.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Generative artificial intelligence (AI) has been playing an important role in various domains. Leveraging its high capability to generate high-fidelity and diverse synthetic data, generative AI is widely applied in diagnostic tasks, such as lung cancer diagnosis using computed tomography (CT). However, existing generative models for lung cancer diagnosis suffer from low efficiency and anatomical imprecision, which limit their clinical applicability. To address these drawbacks, we propose Lung-DDPM+, an improved version of our previous model, Lung-DDPM. This novel approach is a denoising diffusion probabilistic model (DDPM) guided by nodule semantic layouts and accelerated by a pulmonary DPM-solver, enabling the method to focus on lesion areas while achieving a better trade-off between sampling efficiency and quality. Evaluation results on the public LIDC-IDRI dataset suggest that the proposed method achieves 8$ imes$ fewer FLOPs (floating point operations per second), 6.8$ imes$ lower GPU memory consumption, and 14$ imes$ faster sampling compared to Lung-DDPM. Moreover, it maintains comparable sample quality to both Lung-DDPM and other state-of-the-art (SOTA) generative models in two downstream segmentation tasks. We also conducted a Visual Turing Test by an experienced radiologist, showing the advanced quality and fidelity of synthetic samples generated by the proposed method. These experimental results demonstrate that Lung-DDPM+ can effectively generate high-quality thoracic CT images with lung nodules, highlighting its potential for broader applications, such as general tumor synthesis and lesion generation in medical imaging. The code and pretrained models are available at https://github.com/Manem-Lab/Lung-DDPM-PLUS.

Problem

• Improves efficiency in thoracic CT image synthesis
• Enhances anatomical precision in lung cancer diagnosis
• Balances sampling speed and quality in generative models

Innovation

• Denoising diffusion model guided by nodule layouts
• Pulmonary DPM-solver accelerates sampling efficiency
• Generates high-fidelity thoracic CT images efficiently