Accelerating compressive sensing reconstruction OMP algorithm with CPU, GPU, FPGA and domain specific many-core

Kulkarni, Amey

doi:10.1109/iscas.2015.7168797

Cited by 29 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It should be noted that there is still a considerable potential for further accelerating the algorithm. A scalable approach would be to improve the computational ability using a GPU ( 14 ) or FPGA ( 51 ). Second, the parameter values pertaining to the number of iterations, search area, patch size, and spatially encoded factor were set to be equal for ensuring a fair comparison between MI-NLTV and NLTV.…”

Section: Discussionmentioning

confidence: 99%

Mutual Information-Based Non-Local Total Variation Denoiser for Low-Dose Cone-Beam Computed Tomography

et al. 2021

View full text Add to dashboard Cite

Conventional non-local total variation (NLTV) approaches use the weight of a non-local means (NLM) filter, which degrades performance in low-dose cone-beam computed tomography (CBCT) images generated with a low milliampere-seconds (mAs) parameter value because a local patch used to determine the pixel weights comprises noisy-damaged pixels that reduce the similarity between corresponding patches. In this paper, we propose a novel type of NLTV based on a combination of mutual information (MI): MI-NLTV. It is based on a statistical measure for a similarity calculation between the corresponding bins of non-local patches vs. a reference patch. The weight is determined in terms of a statistical measure comprising the MI value between corresponding non-local patches and the reference-patch entropy. The MI-NLTV denoising process is applied to CBCT images generated by the analytical reconstruction algorithm using a ray-driven backprojector (RDB). The MI-NLTV objective function is minimized based on the steepest gradient descent optimization to augment the difference between a real structure and noise, cleaning noisy pixels without significant loss of the fine structure and details that remain in the reconstructed images. The proposed method was evaluated using patient data and actual phantom measurement data acquired with lower mAs. The results show that integrating the RDB further enhances the MI-NLTV denoising-based analytical reconstruction algorithm to achieve a higher CBCT image quality when compared with those generated by NLTV denoising-based approach, with an average of 15.97% higher contrast-to-noise ratio, 2.67% lower root mean square error, 0.12% lower spatial non-uniformity, 1.14% higher correlation, and an average of 18.11% higher detectability index. These quantitative results indicate that the incorporation of MI makes the NLTV more stable and robust than the conventional NLM filter for low-dose CBCT imaging. In addition, achieving clinically acceptable CBCT image quality despite low-mAs projection acquisition can reduce the burden on common online CBCT imaging, improving patient safety throughout the course of radiotherapy.

show abstract

Section: Discussionmentioning

confidence: 99%

Mutual Information-Based Non-Local Total Variation Denoiser for Low-Dose Cone-Beam Computed Tomography

et al. 2021

View full text Add to dashboard Cite

show abstract

“…The recently developed technique "Cloud Computing" would maximize the computational ability without hardware limitation. Another relatively mature technique Field programmable gate array (FPGA) could also achieve a faster processing speed than GPU in DL process [24]. We are working on FPGA realization of SART-DDL and we believe there would be better acceleration solution in the future.…”

Section: Discussionmentioning

confidence: 99%

Low-dose cone-beam CT (LD-CBCT) reconstruction for image-guided radiation therapy (IGRT) by three-dimensional dual-dictionary learning

Song

Zhang

et al. 2020

Radiat Oncol

View full text Add to dashboard Cite

Background: To develop a low-dose cone beam CT (LD-CBCT) reconstruction method named simultaneous algebraic reconstruction technique and dual-dictionary learning (SART-DDL) joint algorithm for image guided radiation therapy (IGRT) and evaluate its imaging quality and clinical application ability. Methods: In this retrospective study, 62 CBCT image sets from February 2018 to July 2018 at west china hospital were randomly collected from 42 head and neck patients (mean [standard deviation] age, 49.7 [11.4] years, 12 females and 30 males). All image sets were retrospectively reconstructed by SART-DDL (resultant D-CBCT image sets) with 18% less clinical raw projections. Reconstruction quality was evaluated by quantitative parameters compared with SART and Total Variation minimization (SART-TV) joint reconstruction algorithm with paired t test. Five-grade subjective grading evaluations were done by two oncologists in a blind manner compared with clinically used Feldkamp-Davis-Kress algorithm CBCT images (resultant F-CBCT image sets) and the grading results were compared by paired Wilcoxon rank test. Registration results between D-CBCT and F-CBCT were compared. D-CBCT image geometry fidelity was tested. Results: The mean peak signal to noise ratio of D-CBCT was 1.7 dB higher than SART-TV reconstructions (P < .001, SART-DDL vs SART-TV, 36.36 ± 0.55 dB vs 34.68 ± 0.28 dB). All D-CBCT images were recognized as clinically acceptable without significant difference with F-CBCT in subjective grading (P > .05). In clinical registration, the maximum translational and rotational difference was 1.8 mm and 1.7 degree respectively. The horizontal, vertical

show abstract

“…But thanks to the independent nature of the OCT dataset, it is possible to incorporate parallel computing to accelerate the processing [51]. Moreover, using a user-programmable Field-programmable gate array (FPGA) might further speed up the CS reconstruction [54]. For the next step, we will explore both pathways and strive to implement real-time OCT image reconstruction and visualization.…”

Section: Towards Real-time Processingmentioning

confidence: 99%

Compressed sensing-enabled phase-sensitive swept-source optical coherence tomography

Ling¹,

Meiniel²,

Singh‐Moon³

et al. 2019

Opt. Express

View full text Add to dashboard Cite

Here we present a novel phase-sensitive swept-source optical coherence tomography (PhS-SS-OCT) system. The simultaneously recorded calibration signal, which is commonly used in SS-OCT to stabilize the phase, is randomly sub-sampled during the acquisition, and it is later reconstructed based on the Compressed Sensing (CS) theory. We first mathematically investigated the method, and verified it through computer simulations. We then conducted a vibrational frequency test and a flow velocity measurement in phantoms to demonstrate the system's capability of handling phase-sensitive tasks. The proposed scheme shows excellent phase stability with greatly discounted data bandwidth compared with conventional procedures. We further showcased the usefulness of the system in biological samples by detecting the blood flow in ex vivo swine left marginal artery. The proposed system is compatible with most of the existing SS-OCT systems and could be a preferred solution for future high-speed phase-sensitive applications.

show abstract

Accelerating compressive sensing reconstruction OMP algorithm with CPU, GPU, FPGA and domain specific many-core

Cited by 29 publications

References 15 publications

Mutual Information-Based Non-Local Total Variation Denoiser for Low-Dose Cone-Beam Computed Tomography

Mutual Information-Based Non-Local Total Variation Denoiser for Low-Dose Cone-Beam Computed Tomography

Low-dose cone-beam CT (LD-CBCT) reconstruction for image-guided radiation therapy (IGRT) by three-dimensional dual-dictionary learning

Compressed sensing-enabled phase-sensitive swept-source optical coherence tomography

Contact Info

Product

Resources

About