Sen Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sen Wang has a total of 42 co-authors. Here are the top co-authors by number of shared publications:
- Adam S. Wang
Department of Radiology, Department of Electrical Engineering, Stanford University, Stanford, CA, USA
9 shared publications Grant M. Stevens
GE Healthcare
6 shared publicationsLi Zhang
6 shared publicationsXiaofei Xu
5 shared publicationsYuxiang Xing
Department of Engineering Physics, Tsinghua University, Room 814A, ZiJing Building 15#, Beijing 100084, Beijing, 100084, CHINA
5 shared publicationsHewei Gao
Engineering Physics, Tsinghua University, Liuqing Bldg #509, Haidian District, Tsinghua University, Bejing, Beijing, 100084, CHINA
4 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (11) , 5336-5348
End-to-End Differentiable Photon Counting CT
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (11) , 5336-5348
End-to-End Differentiable Photon Counting CT
- 1 May 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 13 (03) , 1-14
Deep-learning-based multiple organ segmentation for CT scout images: applications to automatic CT planning
PurposeComputed tomography (CT) scout images are used in CT planning to set the anatomic scan range and optimize the radiation dose. Manual exam planning is highly variable and contributes to excess radiation dose delivered to patients. We present a multi-organ segmentation system for CT scout images with a direct application for automatic scan range prediction, an important component of CT planning.ApproachVariants of a U-Net model with a ResNet-50 encoder were trained to segment 35 distinct anatomic structures from frontal and lateral scout images and evaluated on their performance. Organ boundaries generated by the trained models were used to automatically set the scan range for chest–abdomen–pelvis and abdomen–pelvis CT protocols from frontal and lateral scout images.ResultsThe best-performing models produced organ segmentations with mean Dice scores of 0.823±0.014 for the frontal scout and 0.784±0.013 for the lateral scout. Predicted scan ranges covered ground-truth anatomy for 87% to 100% of test cases, reduced variability in six out of eight tested scan limits, and produced shorter median scan lengths for three out of four scout-to-protocol prediction tasks.ConclusionsThe presented models effectively segment a wide variety of organs from frontal and lateral scout images and demonstrate promising clinical applications to improved scan range prediction and automatic CT planning.
- 1 May 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 13 (03) , 1-14
Deep-learning-based multiple organ segmentation for CT scout images: applications to automatic CT planning
PurposeComputed tomography (CT) scout images are used in CT planning to set the anatomic scan range and optimize the radiation dose. Manual exam planning is highly variable and contributes to excess radiation dose delivered to patients. We present a multi-organ segmentation system for CT scout images with a direct application for automatic scan range prediction, an important component of CT planning.ApproachVariants of a U-Net model with a ResNet-50 encoder were trained to segment 35 distinct anatomic structures from frontal and lateral scout images and evaluated on their performance. Organ boundaries generated by the trained models were used to automatically set the scan range for chest–abdomen–pelvis and abdomen–pelvis CT protocols from frontal and lateral scout images.ResultsThe best-performing models produced organ segmentations with mean Dice scores of 0.823±0.014 for the frontal scout and 0.784±0.013 for the lateral scout. Predicted scan ranges covered ground-truth anatomy for 87% to 100% of test cases, reduced variability in six out of eight tested scan limits, and produced shorter median scan lengths for three out of four scout-to-protocol prediction tasks.ConclusionsThe presented models effectively segment a wide variety of organs from frontal and lateral scout images and demonstrate promising clinical applications to improved scan range prediction and automatic CT planning.
- 1 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (5) , 2455-2467
Implicitly Defined Material Decomposition Estimator and Learned Physics-Informed Neural Proxy for Photon Counting CT
Photon counting detector-based CT (PCCT) systems provide spectral count measurements, enabling material decomposition (MD) for quantitative imaging. Maximum-likelihood estimation (MLE) for MD offers asymptotically unbiased and efficient (minimum variance) results but is usually solved iteratively, making the entire process computationally expensive and time-consuming. Conversely, representative empirical methods relying on calibration aim to construct a direct measurement-decomposition conversion, which can be fast but may suffer from bias or noise amplification. In this work, we show that the iterative MLE method implicitly defines the functional mapping from measurements to estimates, i.e., MD results, and the corresponding mean and noise yield analytical approximation forms from the Implicit Function Theorem. From this perspective, we demonstrate that it is possible to distill knowledge from the implicit function defined by the iterative MLE, i.e., finding the explicit proxy, by leveraging universal approximators such as neural networks and the derivative-aware Sobolev Training paradigm. We show that the proposed method, namely Proxy MD, is both computationally efficient (providing >200 times speedup) and approaches the performance of iterative MLE. Thus it outperforms conventional empirical methods, enabling high-quality real-time quantitative spectral imaging. Furthermore, we also demonstrate that the theoretical Jacobian analysis provides new perspectives in making iterative MD differentiable, enabling differentiable PCCT quantitative imaging and corresponding cross-domain end-to-end training and optimization. The code has been made available at: https://github.com/senwang320/ProxyMD_Demo.
- 1 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (5) , 2455-2467
Implicitly Defined Material Decomposition Estimator and Learned Physics-Informed Neural Proxy for Photon Counting CT
Photon counting detector-based CT (PCCT) systems provide spectral count measurements, enabling material decomposition (MD) for quantitative imaging. Maximum-likelihood estimation (MLE) for MD offers asymptotically unbiased and efficient (minimum variance) results but is usually solved iteratively, making the entire process computationally expensive and time-consuming. Conversely, representative empirical methods relying on calibration aim to construct a direct measurement-decomposition conversion, which can be fast but may suffer from bias or noise amplification. In this work, we show that the iterative MLE method implicitly defines the functional mapping from measurements to estimates, i.e., MD results, and the corresponding mean and noise yield analytical approximation forms from the Implicit Function Theorem. From this perspective, we demonstrate that it is possible to distill knowledge from the implicit function defined by the iterative MLE, i.e., finding the explicit proxy, by leveraging universal approximators such as neural networks and the derivative-aware Sobolev Training paradigm. We show that the proposed method, namely Proxy MD, is both computationally efficient (providing >200 times speedup) and approaches the performance of iterative MLE. Thus it outperforms conventional empirical methods, enabling high-quality real-time quantitative spectral imaging. Furthermore, we also demonstrate that the theoretical Jacobian analysis provides new perspectives in making iterative MD differentiable, enabling differentiable PCCT quantitative imaging and corresponding cross-domain end-to-end training and optimization. The code has been made available at: https://github.com/senwang320/ProxyMD_Demo.
- 12 December 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (24) , 245011
Scout-Dose-TCM: direct and prospective scout-based estimation of personalized organ and effective doses from tube current modulated CT exams
Objective. This study proposes Scout-Dose-TCM for direct, prospective estimation of organ-level and effective doses under tube current modulation (TCM) and compares its performance with two established methods. Approach. Contrast-enhanced chest-abdomen-pelvis CT exams from 130 adults (120 kVp, TCM) were analyzed. Reference organ doses for six organs (lungs, kidneys, liver, pancreas, bladder, spleen) were calculated using Monte Carlo (MC)-GPU and TotalSegmentator. These data trained Scout-Dose-TCM, a deep-learning model that predicts organ-level doses corresponding to discrete cosine transform (DCT) basis functions, enabling real-time estimation for any TCM profile. The model includes a feature-learning module that extracts contextual information from lateral and frontal scouts and scan range, and a dose-learning module that outputs DCT-based dose estimates. The loss function incorporated the DCT formulation, ensuring accurate predictions across variable TCM patterns. For comparison, dose estimation was performed following AAPM Task Group 204 (Global CTDI vol ) and its TCM-adapted and organ-specific version (Organ CTDI vol ). The three evaluated methods were extended to estimate effective dose and compared against the approach in AAPM Report 96, which applies a fixed dose-length product-to-effective-dose conversion factor. Five-fold cross-validation assessed generalizability via mean absolute percentage dose errors and R 2 correlations with MC benchmarks. Main Results. Mean organ-level absolute percentage errors were 13% (Global CTDI vol ), 9% (Organ CTDI vol ), and 7% (Scout-Dose-TCM). The largest discrepancies occurred for the bladder (15%, 13%, and 9%). Scout-Dose-TCM significantly reduced organ-level dose errors versus Global CTDI vol ( p < 1 × 10 −7 ) and improved predictions for liver, bladder, and pancreas versus Organ CTDI vol ( p ⩽ 0.005). It also achieved higher R 2 values and lower effective-dose error (3.01%) than Global (3.87%) and Organ (4.16%) CTDI vol , and the AAPM Task Group 96 method (31%). Significance. Scout-Dose-TCM outperformed Global CTDI vol and was comparable to or better than Organ CTDI vol for organ and effective dose estimation without requiring segmentations at inference, demonstrating its clinical potential for prospective dose estimation in CT.
- 12 December 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (24) , 245011
Scout-Dose-TCM: direct and prospective scout-based estimation of personalized organ and effective doses from tube current modulated CT exams
Objective. This study proposes Scout-Dose-TCM for direct, prospective estimation of organ-level and effective doses under tube current modulation (TCM) and compares its performance with two established methods. Approach. Contrast-enhanced chest-abdomen-pelvis CT exams from 130 adults (120 kVp, TCM) were analyzed. Reference organ doses for six organs (lungs, kidneys, liver, pancreas, bladder, spleen) were calculated using Monte Carlo (MC)-GPU and TotalSegmentator. These data trained Scout-Dose-TCM, a deep-learning model that predicts organ-level doses corresponding to discrete cosine transform (DCT) basis functions, enabling real-time estimation for any TCM profile. The model includes a feature-learning module that extracts contextual information from lateral and frontal scouts and scan range, and a dose-learning module that outputs DCT-based dose estimates. The loss function incorporated the DCT formulation, ensuring accurate predictions across variable TCM patterns. For comparison, dose estimation was performed following AAPM Task Group 204 (Global CTDI vol ) and its TCM-adapted and organ-specific version (Organ CTDI vol ). The three evaluated methods were extended to estimate effective dose and compared against the approach in AAPM Report 96, which applies a fixed dose-length product-to-effective-dose conversion factor. Five-fold cross-validation assessed generalizability via mean absolute percentage dose errors and R 2 correlations with MC benchmarks. Main Results. Mean organ-level absolute percentage errors were 13% (Global CTDI vol ), 9% (Organ CTDI vol ), and 7% (Scout-Dose-TCM). The largest discrepancies occurred for the bladder (15%, 13%, and 9%). Scout-Dose-TCM significantly reduced organ-level dose errors versus Global CTDI vol ( p < 1 × 10 −7 ) and improved predictions for liver, bladder, and pancreas versus Organ CTDI vol ( p ⩽ 0.005). It also achieved higher R 2 values and lower effective-dose error (3.01%) than Global (3.87%) and Organ (4.16%) CTDI vol , and the AAPM Task Group 96 method (31%). Significance. Scout-Dose-TCM outperformed Global CTDI vol and was comparable to or better than Organ CTDI vol for organ and effective dose estimation without requiring segmentations at inference, demonstrating its clinical potential for prospective dose estimation in CT.
- 8 April 2025
- Published by SPIE-Intl Soc Optical Eng
- Vol. 13405, 16-100
Implicitly defined PCCT material decomposition estimator and learned physics-informed neural proxy
Photon counting detector-based CT systems provide spectral count measurements, enabling material decomposition for quantitative imaging. Maximum-likelihood estimation (MLE) for material decomposition offers asymptotically unbiased and efficient (minimum variance) results but is usually solved iteratively, making the entire process computationally expensive and time-consuming. Conversely, representative empirical methods relying on grid calibration aim to construct a direct measurement-decomposition conversion, which can be fast but may suffer from bias or noise amplification. In this work, we show that the iterative MLE method implicitly defines the functional mapping from measurements to estimates, i.e. material decomposition results, and the corresponding mean and noise yield analytical approximation forms from the Implicit Function Theorem. From this perspective, we demonstrate that it is possible to distill knowledge from the implicit function defined by the iterative MLE, i.e., finding the explicit proxy, by leveraging universal approximators such as neural networks and the derivative-aware Sobolev Learning paradigm. We show that the proposed method, namely proxy MD, is both computationally efficient (providing >3000 times speedup in numerical simulation) and approaches the performance of iterative MLE. Thus, it outperforms conventional empirical methods and enables high-quality real-time quantitative spectral imaging.
- 8 April 2025
- Published by SPIE-Intl Soc Optical Eng
- Vol. 13405, 16-100
Implicitly defined PCCT material decomposition estimator and learned physics-informed neural proxy
Photon counting detector-based CT systems provide spectral count measurements, enabling material decomposition for quantitative imaging. Maximum-likelihood estimation (MLE) for material decomposition offers asymptotically unbiased and efficient (minimum variance) results but is usually solved iteratively, making the entire process computationally expensive and time-consuming. Conversely, representative empirical methods relying on grid calibration aim to construct a direct measurement-decomposition conversion, which can be fast but may suffer from bias or noise amplification. In this work, we show that the iterative MLE method implicitly defines the functional mapping from measurements to estimates, i.e. material decomposition results, and the corresponding mean and noise yield analytical approximation forms from the Implicit Function Theorem. From this perspective, we demonstrate that it is possible to distill knowledge from the implicit function defined by the iterative MLE, i.e., finding the explicit proxy, by leveraging universal approximators such as neural networks and the derivative-aware Sobolev Learning paradigm. We show that the proposed method, namely proxy MD, is both computationally efficient (providing >3000 times speedup in numerical simulation) and approaches the performance of iterative MLE. Thus, it outperforms conventional empirical methods and enables high-quality real-time quantitative spectral imaging.
- 8 April 2025
- Published by SPIE-Intl Soc Optical Eng
- Vol. 13405, 495-498
Multiple organ segmentation for CT scout images
CT scout images are used to plan the anatomic scan range and optimize radiation dose. This initial step customizes CT examinations to the individual patient and maximizes image quality while minimizing radiation exposure. For these exams to be fully personalized at the organ level, the precise location of the patient’s relevant anatomy within the CT scan range is of utmost importance. In this work, we investigate methods for the automatic segmentation of multiple organs from CT scout images. We use variations of a U-Net model with a ResNet-50 encoder backbone pretrained on ImageNet-1K, and examine their segmentation performance for frontal and lateral scout views. We demonstrate that automatic organ segmentation for scout images is a highly feasible and attractive task, with applications extending beyond CT exam planning to include radiograph-based surgical planning and guiding organ-specific procedures.
- 8 April 2025
- Published by SPIE-Intl Soc Optical Eng
- Vol. 13405, 495-498
Multiple organ segmentation for CT scout images
CT scout images are used to plan the anatomic scan range and optimize radiation dose. This initial step customizes CT examinations to the individual patient and maximizes image quality while minimizing radiation exposure. For these exams to be fully personalized at the organ level, the precise location of the patient’s relevant anatomy within the CT scan range is of utmost importance. In this work, we investigate methods for the automatic segmentation of multiple organs from CT scout images. We use variations of a U-Net model with a ResNet-50 encoder backbone pretrained on ImageNet-1K, and examine their segmentation performance for frontal and lateral scout views. We demonstrate that automatic organ segmentation for scout images is a highly feasible and attractive task, with applications extending beyond CT exam planning to include radiograph-based surgical planning and guiding organ-specific procedures.
- 14 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging
Low-dose fluoroscopy is essential for real-time X-ray visualization, supporting dynamic diagnostic assessments while minimizing harmful radiation exposure to patients. However, low-dose imaging introduces noise that can impair diagnostic accuracy. Although numerous deep learning methods have been developed for noise reduction in medical imaging, the unique challenges of fluoroscopy—such as motion artifacts due to its real-time nature, limited access to clean reference data, high noise levels—diminish the effectiveness of current deep learning-based denoising techniques, leaving research in this area relatively constrained. To address these challenges, we present three key innovations. First, we propose an unsupervised framework for dynamic, context-aware denoising in fluoroscopy, introducing the multi-scale recurrent attention U-Net (MSR2AU-Net) to effectively reduce noise without clean data by directly targeting initial noise. Second, our proposed dual noise suppression strategy combines a knowledge distillation-based module for uncorrelated noise with a recursive filtering module for correlated noise, enhancing both denoising quality and motion stability. Finally, we design a pixel-wise dynamic object motion cross-fusion matrix combined with an edge-preserving loss function to preserve fine details amidst motion changes. Our model was evaluated on 3,500 fluoroscopy images from dynamic phantoms (2,400 for training and 1,100 for testing) and 350 clinical images from spinal surgery patients, confirming its effectiveness in clinical settings. To further validate the robustness of our approach, we also tested it on the “2016 NIH-AAPM-Mayo Clinic Low Dose CT Grand Challenge” dataset, using 4,800 images for training and 1,136 for testing, demonstrating efficacy across both fluoroscopy and CT imaging. Quantitative evaluations demonstrated that our approach outperformed existing unsupervised algorithms, achieving a Structural Similarity Index Measure (SSIM) of 0.9803 and a Peak Signal-to-Noise Ratio (PSNR) of 39.12 dB on the dynamic phantom dataset, and an SSIM of 0.9591 with a PSNR of 36.62 dB on the Mayo CT dataset. These results indicate that our approach surpasses state-of-the-art unsupervised algorithms in both visual quality and quantitative metrics, achieving performance comparable to well-established supervised methods in low-dose fluoroscopy and CT imaging. The source code will be available at https://github.com/sunyoungIT/UDCA-Net.git.
- 14 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging
Low-dose fluoroscopy is essential for real-time X-ray visualization, supporting dynamic diagnostic assessments while minimizing harmful radiation exposure to patients. However, low-dose imaging introduces noise that can impair diagnostic accuracy. Although numerous deep learning methods have been developed for noise reduction in medical imaging, the unique challenges of fluoroscopy—such as motion artifacts due to its real-time nature, limited access to clean reference data, high noise levels—diminish the effectiveness of current deep learning-based denoising techniques, leaving research in this area relatively constrained. To address these challenges, we present three key innovations. First, we propose an unsupervised framework for dynamic, context-aware denoising in fluoroscopy, introducing the multi-scale recurrent attention U-Net (MSR2AU-Net) to effectively reduce noise without clean data by directly targeting initial noise. Second, our proposed dual noise suppression strategy combines a knowledge distillation-based module for uncorrelated noise with a recursive filtering module for correlated noise, enhancing both denoising quality and motion stability. Finally, we design a pixel-wise dynamic object motion cross-fusion matrix combined with an edge-preserving loss function to preserve fine details amidst motion changes. Our model was evaluated on 3,500 fluoroscopy images from dynamic phantoms (2,400 for training and 1,100 for testing) and 350 clinical images from spinal surgery patients, confirming its effectiveness in clinical settings. To further validate the robustness of our approach, we also tested it on the “2016 NIH-AAPM-Mayo Clinic Low Dose CT Grand Challenge” dataset, using 4,800 images for training and 1,136 for testing, demonstrating efficacy across both fluoroscopy and CT imaging. Quantitative evaluations demonstrated that our approach outperformed existing unsupervised algorithms, achieving a Structural Similarity Index Measure (SSIM) of 0.9803 and a Peak Signal-to-Noise Ratio (PSNR) of 39.12 dB on the dynamic phantom dataset, and an SSIM of 0.9591 with a PSNR of 36.62 dB on the Mayo CT dataset. These results indicate that our approach surpasses state-of-the-art unsupervised algorithms in both visual quality and quantitative metrics, achieving performance comparable to well-established supervised methods in low-dose fluoroscopy and CT imaging. The source code will be available at https://github.com/sunyoungIT/UDCA-Net.git.
- 29 January 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (3) , 035014
Automated estimation of individualized organ-specific dose and noise from clinical CT scans
Objective. Radiation dose and diagnostic image quality are opposing constraints in x-ray computed tomography (CT). Conventional methods do not fully account for organ-level radiation dose and noise when considering radiation risk and clinical task. In this work, we develop a pipeline to generate individualized organ-specific dose and noise at desired dose levels from clinical CT scans. Approach. To estimate organ-specific dose and noise, we compute dose maps, noise maps at desired dose levels and organ segmentations. In our pipeline, dose maps are generated using Monte Carlo simulation. The noise map is obtained by scaling the inserted noise in synthetic low-dose emulation in order to avoid anatomical structures, where the scaling coefficients are empirically calibrated. Organ segmentations are generated by a deep learning-based method (TotalSegmentator). The proposed noise model is evaluated on a clinical dataset of 12 CT scans, a phantom dataset of 3 uniform phantom scans, and a cross-site dataset of 26 scans. The accuracy of deep learning-based segmentations for organ-level dose and noise estimates was tested using a dataset of 41 cases with expert segmentations of six organs: lungs, liver, kidneys, bladder, spleen, and pancreas. Main results. The empirical noise model performs well, with an average RMSE approximately 1.5 HU and an average relative RMSE approximately 5% across different dose levels. The segmentation from TotalSegmentator yielded a mean Dice score of 0.8597 across the six organs (max = 0.9315 in liver, min = 0.6855 in pancreas). The resulting error in organ-level dose and noise estimation was less than 2% for most organs. Significance. The proposed pipeline can output individualized organ-specific dose and noise estimates accurately for personalized protocol evaluation and optimization. It is fully automated and can be scalable to large clinical datasets. This pipeline can be used to optimize image quality for specific organs and thus clinical tasks, without adversely affecting overall radiation dose.
- 29 January 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (3) , 035014
Automated estimation of individualized organ-specific dose and noise from clinical CT scans
Objective. Radiation dose and diagnostic image quality are opposing constraints in x-ray computed tomography (CT). Conventional methods do not fully account for organ-level radiation dose and noise when considering radiation risk and clinical task. In this work, we develop a pipeline to generate individualized organ-specific dose and noise at desired dose levels from clinical CT scans. Approach. To estimate organ-specific dose and noise, we compute dose maps, noise maps at desired dose levels and organ segmentations. In our pipeline, dose maps are generated using Monte Carlo simulation. The noise map is obtained by scaling the inserted noise in synthetic low-dose emulation in order to avoid anatomical structures, where the scaling coefficients are empirically calibrated. Organ segmentations are generated by a deep learning-based method (TotalSegmentator). The proposed noise model is evaluated on a clinical dataset of 12 CT scans, a phantom dataset of 3 uniform phantom scans, and a cross-site dataset of 26 scans. The accuracy of deep learning-based segmentations for organ-level dose and noise estimates was tested using a dataset of 41 cases with expert segmentations of six organs: lungs, liver, kidneys, bladder, spleen, and pancreas. Main results. The empirical noise model performs well, with an average RMSE approximately 1.5 HU and an average relative RMSE approximately 5% across different dose levels. The segmentation from TotalSegmentator yielded a mean Dice score of 0.8597 across the six organs (max = 0.9315 in liver, min = 0.6855 in pancreas). The resulting error in organ-level dose and noise estimation was less than 2% for most organs. Significance. The proposed pipeline can output individualized organ-specific dose and noise estimates accurately for personalized protocol evaluation and optimization. It is fully automated and can be scalable to large clinical datasets. This pipeline can be used to optimize image quality for specific organs and thus clinical tasks, without adversely affecting overall radiation dose.
- 28 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 44 (1) , 530-542
Emulating Low-Dose PCCT Image Pairs With Independent Noise for Self-Supervised Spectral Image Denoising
Photon counting CT (PCCT) acquires spectral measurements and enables generation of material decomposition (MD) images that provide distinct advantages in various clinical situations. However, noise amplification is observed in MD images, and denoising is typically applied. Clean or high-quality references are rare in clinical scans, often making supervised learning (Noise2Clean) impractical. Noise2Noise is a self-supervised counterpart, using noisy images and corresponding noisy references with zero-mean, independent noise. PCCT counts transmitted photons separately, and raw measurements are assumed to follow a Poisson distribution in each energy bin, providing the possibility to create noise-independent pairs. The approach is to use binomial selection to split the counts into two low-dose scans with independent noise. We prove that the reconstructed spectral images inherit the noise independence from counts domain through noise propagation analysis and also validated it in numerical simulation and experimental phantom scans. The method offers the flexibility to split measurements into desired dose levels while ensuring the reconstructed images share identical underlying features, thereby strengthening the model’s robustness for input dose levels and capability of preserving fine details. In both numerical simulation and experimental phantom scans, we demonstrated that Noise2Noise with binomial selection outperforms other common self-supervised learning methods based on different presumptive conditions.
- 28 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 44 (1) , 530-542
Emulating Low-Dose PCCT Image Pairs With Independent Noise for Self-Supervised Spectral Image Denoising
Photon counting CT (PCCT) acquires spectral measurements and enables generation of material decomposition (MD) images that provide distinct advantages in various clinical situations. However, noise amplification is observed in MD images, and denoising is typically applied. Clean or high-quality references are rare in clinical scans, often making supervised learning (Noise2Clean) impractical. Noise2Noise is a self-supervised counterpart, using noisy images and corresponding noisy references with zero-mean, independent noise. PCCT counts transmitted photons separately, and raw measurements are assumed to follow a Poisson distribution in each energy bin, providing the possibility to create noise-independent pairs. The approach is to use binomial selection to split the counts into two low-dose scans with independent noise. We prove that the reconstructed spectral images inherit the noise independence from counts domain through noise propagation analysis and also validated it in numerical simulation and experimental phantom scans. The method offers the flexibility to split measurements into desired dose levels while ensuring the reconstructed images share identical underlying features, thereby strengthening the model’s robustness for input dose levels and capability of preserving fine details. In both numerical simulation and experimental phantom scans, we demonstrated that Noise2Noise with binomial selection outperforms other common self-supervised learning methods based on different presumptive conditions.
- 25 July 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (S1) , S12805-14
Spectral optimization using fast kV switching and filtration for photon counting CT with realistic detector responses: a simulation study
PurposePhoton counting CT (PCCT) provides spectral measurements for material decomposition. However, the image noise (at a fixed dose) depends on the source spectrum. Our study investigates the potential benefits from spectral optimization using fast kV switching and filtration to reduce noise in material decomposition.ApproachThe effect of the input spectra on noise performance in both two-basis material decomposition and three-basis material decomposition was compared using Cramer-Rao lower bound analysis in the projection domain and in a digital phantom study in the image domain. The fluences of different spectra were normalized using the CT dose index to maintain constant dose levels. Four detector response models based on Si or CdTe were included in the analysis.ResultsFor single kV scans, kV selection can be optimized based on the imaging task and object size. Furthermore, our results suggest that noise in material decomposition can be substantially reduced with fast kV switching. For two-material decomposition, fast kV switching reduces the standard deviation (SD) by ∼10%. For three-material decomposition, greater noise reduction in material images was found with fast kV switching (26.2% for calcium and 25.8% for iodine, in terms of SD), which suggests that challenging tasks benefit more from the richer spectral information provided by fast kV switching.ConclusionsThe performance of PCCT in material decomposition can be improved by optimizing source spectrum settings. Task-specific tube voltages can be selected for single kV scans. Also, our results demonstrate that utilizing fast kV switching can substantially reduce the noise in material decomposition for both two- and three-material decompositions, and a fixed Gd filter can further enhance such improvements for two-material decomposition.
- 25 July 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (S1) , S12805-14
Spectral optimization using fast kV switching and filtration for photon counting CT with realistic detector responses: a simulation study
PurposePhoton counting CT (PCCT) provides spectral measurements for material decomposition. However, the image noise (at a fixed dose) depends on the source spectrum. Our study investigates the potential benefits from spectral optimization using fast kV switching and filtration to reduce noise in material decomposition.ApproachThe effect of the input spectra on noise performance in both two-basis material decomposition and three-basis material decomposition was compared using Cramer-Rao lower bound analysis in the projection domain and in a digital phantom study in the image domain. The fluences of different spectra were normalized using the CT dose index to maintain constant dose levels. Four detector response models based on Si or CdTe were included in the analysis.ResultsFor single kV scans, kV selection can be optimized based on the imaging task and object size. Furthermore, our results suggest that noise in material decomposition can be substantially reduced with fast kV switching. For two-material decomposition, fast kV switching reduces the standard deviation (SD) by ∼10%. For three-material decomposition, greater noise reduction in material images was found with fast kV switching (26.2% for calcium and 25.8% for iodine, in terms of SD), which suggests that challenging tasks benefit more from the richer spectral information provided by fast kV switching.ConclusionsThe performance of PCCT in material decomposition can be improved by optimizing source spectrum settings. Task-specific tube voltages can be selected for single kV scans. Also, our results demonstrate that utilizing fast kV switching can substantially reduce the noise in material decomposition for both two- and three-material decompositions, and a fixed Gd filter can further enhance such improvements for two-material decomposition.