BS-LDM: Effective Bone Suppression in High-Resolution Chest X-Ray Images with Conditional Latent Diffusion Models

Yifei Sun1, Zhanghao Chen1, Hao Zheng1, Wenming Deng2, Jin Liu3, Wenwen Min4, Ahmed Elazab5, Xiang Wan6, Changmiao Wang6*, Ruiquan Ge1*
1 Hangzhou Dianzi University, 2 National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital&Shenzhen Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 3 Central South University, 4 Yunnan University, 5 Shenzhen University, 6 Shenzhen Research Institute of Big Data
* Indicates Corresponding Author
Paper Code Hugging Face Hugging Face arXiv

What is bone suppression?

🔻 Try hovering the mouse over the image below and moving it horizontally!

The left side is a Chest X-Ray (CXR) image while the right side is a soft tissue image obtained by bone suppression using our BS-LDM model.

Bone suppression in CXR imaging refers to computational or imaging techniques that minimize or remove the visibility of bone structures from CXR images. The goal is to enhance the clarity of underlying soft tissues, particularly the lungs, to improve diagnostic accuracy for pulmonary conditions.

Abstract

Lung diseases represent a significant global health challenge, with Chest X-Ray (CXR) being a key diagnostic tool due to their accessibility and affordability. Nonetheless, the detection of pulmonary lesions is often hindered by overlapping bone structures in CXR images, leading to potential misdiagnoses. To address this issue, we developed an end-to-end framework called BS-LDM, designed to effectively suppress bone in high-resolution CXR images. This framework is based on conditional latent diffusion models and incorporates a multi-level hybrid loss-constrained vector-quantized generative adversarial network which is crafted for perceptual compression, ensuring the preservation of details. To further enhance the framework's performance, we introduce offset noise and a temporal adaptive thresholding strategy. These additions help minimize discrepancies in generating low-frequency information, thereby improving the clarity of the generated soft tissue images. Additionally, we have compiled a high-quality bone suppression dataset named SZCH-X-Rays. This dataset includes 818 pairs of high-resolution CXR and dual-energy subtraction soft tissue images collected from a partner hospital. Moreover, we processed 241 data pairs from the JSRT dataset into negative images, which are more commonly used in clinical practice. Our comprehensive experiments and downstream evaluations reveal that BS-LDM excels in bone suppression, underscoring its significant clinical value.


Method

MY ALT TEXT

The overview framework of our proposed BS-LDM.


MY ALT TEXT

Visualization of high-frequency and low-frequency feature decomposition of latent variables before and after Gaussian noise addition using Discrete Fourier Transform. The results are pseudo-colored for ease of demonstration.


MY ALT TEXT

Illustration of the composition of offset noise.


Highlights

🌟 A LDM-based framework for high-resolution bone suppression in CXRs, and introduce ML-VQGAN for effective perceptual compression and detail retention.

🌟 Incorporate offset noise and a temporal adaptive thresholding strategy to help minimize discrepancies in low-frequency information and enhance the quality of generated images.

🌟 Our comprehensive experiments, as well as clinical and automated downstream evaluations demonstrated excellent image quality and substantial diagnostic improvements, underscoring the clinical significance of our work.


Experiments

MY ALT TEXT

Visulization of different methods comparison on SZCH-X-Rays and JSRT datasets.

MY ALT TEXT Visualization of ablation studies on BS-LDM. The upper row shows generated soft tissue images under different conditions, while the lower row presents corresponding pixel intensity histograms, where rightward shift indicates brighter images, and wider spread indicates higher contrast.

MY ALT TEXT Automated downstream evaluation of BS-LDM on Shenzhen chest X-ray using three classification models.

Clinical Evaluation of Image Quality Assessment

MY ALT TEXT

Image quality assessment of BS-LDM scored by three radiologists with various levels of experience.

Clinical Evaluation of Diagnostic Utility Assessment

MY ALT TEXT

Diagnostic utility assessment of BS-LDM calculated from the diagnostic results of two radiologists with 6 and 11 years of experience, respectively.

BibTeX

@article{sun2024bsldmeffectivebonesuppression,
      title={BS-LDM: Effective Bone Suppression in High-Resolution Chest X-Ray Images with Conditional Latent Diffusion Models}, 
      author={Yifei Sun and Zhanghao Chen and Hao Zheng and Wenming Deng and Jin Liu and Wenwen Min and Ahmed Elazab and Xiang Wan and Changmiao Wang and Ruiquan Ge},
      year={2024},
      eprint={2412.15670},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2412.15670}, 
}