According to American Cancer Society statistics, lung cancer is the leading cause of cancer-related death in the United States, 212,380 of new cases diagnosed and 160,390 deaths in 2007. Early detection of lung cancers (less than 3 cm) can achieve a 90% ten-year survival rate. Early sign of the cancer is small lung nodules. Current screening of the lung nodules is performed by high-resolution computed tomography (CT), which carries a significant radiation and could increase the risk of getting cancer by as high as 2% according to a recent report in The New England Journal of Medicine. In addition to the screening, more CT scans are performed for follow-up and/or biopsy procedures. Reducing the radiation risk has been attempted by CT manufacturers by both hardware optimization and software enhancement. We have been exploring adaptive noise-treatment strategies to reconstruct similar image quality at significantly low mAs level for ultra low-dose CT applications on currently available hardware configuration. Iterative image reconstruction under a statistical cost function is one of the strategies which needs powerful computing engine (costs more than a half million dollars). Analytical image reconstruction after data restoration by a statistical cost function is another strategy which generates similar results as the iterative means with a dramatic reduction of computing burden. Our pilot studies by both phantom and volunteer experiments have demonstrated great potential of the latter restoration strategy for radiation reduction while retaining the image quality and reconstruction speed on currently available CT scanners. The proposed specific aims to further explore the potential for screening lung nodules are: (SA-1). To further investigate the adaptive noise-treatment strategies toward as low mAs as achievable for lung screening: Because the first and second moments of low-mAs CT data contain the essential statistical information about the noise (higher order moments have less impact on noise reduction), we will study the properties of sample mean and variance of the data as mAs level goes down as low as achievable. In addition, data correlations in the three-dimensional spatial domain associated with tomographic imaging will be investigated. Both the noise properties and data correlation will be incorporated into a statistical cost function, i.e., Kharhunen-Lohve domain penalized weighted least-squares, which can be efficiently minimized for data restoration by an analytical fashion at the highest speed. Image reconstruction from the restored data will also be analytical at the highest speed. For comparison purpose, iterative image reconstruction under a similar statistical cost function will be refined. (SA-2). To evaluate the investigated adaptive strategies by the detection of small lung nodules: The presented strategies will be first evaluated by repeated experiments on anthropomorphic phantoms with variable low mAs protocols using noise-resolution tradeoff measure and receiver operating characteristics (ROC) and channelized Hotelling trace (CHT) observer studies. Then the evaluation will be on patient lung nodule detection with comparison to currently-used mAs level by a same CT scanner, where quantitative measures will be made using performance equivalence tests and ROC studies. The successfully evaluated strategies may lead to a large clinical trial for ultra low-dose CT screening of the lung nodules, and could be extended to screening of other vital organs, such as the colon, heart, and breasts.

Public Health Relevance

Current practice of computed tomography (CT) in clinic frequently delivers excessive X-ray radiation to the patients by using a higher mAs scanning protocol than needed. This causes a major concern when screening is the clinical task, e.g., screening lung cancer. If the mAs value is lowered, image noise will increase and streak artifacts may present (because there is no effective noise treatment in current CT scanners), compromising the clinical assessment. This proposal aims to reduce the X-ray exposure risk by lowering the mAs value as low as achievable, while retaining the image quality suitable to the clinical task. The key technical component is a software module which can be easily adapted by current clinical CT scanners without any hardware modification except for a few seconds of computing time. The module reads in CT data, analyzes and then treats the data noise prior to reconstructing the data, preventing image noise and artifact.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Research Project (R01)
Project #
Application #
Study Section
Biomedical Imaging Technology Study Section (BMIT)
Program Officer
Tandon, Pushpa
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
State University New York Stony Brook
Schools of Medicine
Stony Brook
United States
Zip Code
Chen, Bo; Bian, Zhaoying; Zhou, Xiaohui et al. (2018) A new Mumford-Shah total variation minimization based model for sparse-view x-ray computed tomography image reconstruction. Neurocomputing 285:74-81
Chen, Wensheng; You, Jie; Pan, Binbin et al. (2018) A sparse representation and dictionary learning based algorithm for image restoration in the presence of Rician noise. Neurocomputing 286:130-140
Zhang, Hao; Ma, Jianhua; Wang, Jing et al. (2017) Assessment of prior image induced nonlocal means regularization for low-dose CT reconstruction: Change in anatomy. Med Phys 44:e264-e278
Zhang, Houjin; Zeng, Dong; Lin, Jiahui et al. (2017) Iterative reconstruction for dual energy CT with an average image-induced nonlocal means regularization. Phys Med Biol 62:5556-5574
Zhang, Hao; Zeng, Dong; Zhang, Hua et al. (2017) Applications of nonlocal means algorithm in low-dose X-ray CT image processing and reconstruction: A review. Med Phys 44:1168-1185
Zhang, Xi; Xu, Xiaopan; Tian, Qiang et al. (2017) Radiomics assessment of bladder cancer grade using texture features from diffusion-weighted imaging. J Magn Reson Imaging 46:1281-1288
Niu, Shanzhou; Zhang, Shanli; Huang, Jing et al. (2016) Low-dose cerebral perfusion computed tomography image restoration via low-rank and total variation regularizations. Neurocomputing 197:143-160
Zhang, Hao; Han, Hao; Liang, Zhengrong et al. (2016) Extracting Information From Previous Full-Dose CT Scan for Knowledge-Based Bayesian Reconstruction of Current Low-Dose CT Images. IEEE Trans Med Imaging 35:860-70
Zeng, Dong; Gong, Changfei; Bian, Zhaoying et al. (2016) Robust dynamic myocardial perfusion CT deconvolution for accurate residue function estimation via adaptive-weighted tensor total variation regularization: a preclinical study. Phys Med Biol 61:8135-8156
Han, Hao; Lin, Qin; Li, Lihong et al. (2016) ?-Information-Based Registration of Dynamic Scans for Magnetic Resonance Cystography. IEEE J Biomed Health Inform 20:1160-70

Showing the most recent 10 out of 40 publications