Related Experiment Video
Updated: Jul 28, 2025
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
Direct attenuation correction for 99mTc-3PRGD2 chest SPECT lung cancer images using deep learning
Haiqun Xing1, Tong Wang1, Xiaona Jin1
1Department of Nuclear Medicine, Peking Union Medical College Hospital, Chinese Academy of Medical Science and Peking Union Medical College, Beijing Key Laboratory of Molecular Targeted Diagnosis and Therapy in Nuclear Medicine, Beijing, China.
Deep learning accurately corrects SPECT images for lung cancer diagnosis using 99mTc-3PRGD2. This novel method offers a feasible alternative for attenuation correction, especially when CT is unavailable, aiding treatment evaluation.
Area of Science:
- Medical Imaging
- Radiochemistry
- Artificial Intelligence
Background:
- Attenuation correction is crucial for accurate single photon emission computed tomography (SPECT) imaging in lung cancer diagnosis and treatment monitoring.
- 99mTc-3PRGD2 is a promising radiotracer for early lung cancer detection and evaluating treatment efficacy.
- Developing advanced methods for SPECT image correction is essential for improving diagnostic accuracy.
Purpose of the Study:
- To investigate the feasibility and accuracy of a deep learning-based method for direct attenuation correction of 99mTc-3PRGD2 chest SPECT images.
- To compare the performance of deep learning attenuation correction (DL-AC) with conventional CT-based attenuation correction (CT-AC).
- To evaluate the potential of DL-AC for lung cancer diagnosis and treatment effect assessment.
Main Methods:
- A retrospective analysis of 53 lung cancer patients who underwent 99mTc-3PRGD2 SPECT/CT imaging.
- Development of a deep learning model (3D Unet) trained on CT-AC images as the ground truth to perform direct attenuation correction on non-attenuation corrected (NAC) SPECT images.
- Evaluation of the DL-AC model using quantitative metrics (MAE, MSE, PSNR, SSIM, NRMSE, NMI) and analysis of tumor-to-background ratios in lung lesions on a dedicated testing set.
Main Results:
- The deep learning attenuation correction (DL-AC) method achieved high accuracy, with key imaging quality metrics comparable to CT-AC.
- Specific metrics for DL-AC included PSNR > 42, SSIM > 0.8, and NRMSE < 0.11.
- No significant difference was observed in the tumor-to-background ratios of lung lesions between the CT-AC and DL-AC groups (p = 0.81).
Conclusions:
- The deep learning-based attenuation correction method is highly accurate and feasible for 99mTc-3PRGD2 chest SPECT imaging.
- This DL-AC approach provides a viable alternative for SPECT attenuation correction, particularly in scenarios where CT is not available.
- The findings support the use of DL-AC for improved lung cancer diagnosis and treatment effect evaluation using SPECT/CT scans.
