Deep learning based correction of low performing pixel in computed tomography

Low Performing Pixel (LPP)/bad pixel in CT detectors cause ring and streaks artifacts, structured non-uniformities and deterioration of the image quality. These artifacts make the image unusable for diagnostic purposes. A missing/defective detector pixel translates to a channel missing across all views in sinogram domain and its effect gets spill over entire image in reconstruction domain as artifacts. Most of the existing ring and streak removal algorithms perform correction only in the reconstructed image domain. In this work, we propose a supervised deep learning algorithm that operates in sinogram domain to remove distortions cause by the LPP. This method leverages CT scan geometry, including conjugate ray information to learn the interpolation in sinogram domain. While the experiments are designed to cover the entire detector space, we emphasize on LPPs near detector iso-center as these have most adverse impact on image quality specially if the LPPs fall on the high frequency region (bone-tissue interface). We demonstrated efficacy of the proposed method using data acquired on GE RevACT multi-slice CT system with flat-panel detector. Experimental results on head scans show significant reduction in ring artifacts regardless of LPP location in the detector geometry. We have simulated isolated LPPs accounting for 5% and 10% of total channels. Detailed statistical analysis illustrates approximately 5dB improvement in SNR in both sinogram and reconstruction domain as compared to classical bicubic and Lagrange interpolation methods. Also, with reduction in ring and streak artifacts, the perceptual image quality is improved across all the test images.

Volumetric CT with sparse detector arrays (and application to Si-strip photon counters)

Novel x-ray medical imaging sensors, such as photon counting detectors (PCDs) and large area CCD and CMOS cameras can involve irregular and/or sparse sampling of the detector plane. Application of such detectors to CT involves undersampling that is markedly different from the commonly considered case of sparse angular sampling. This work investigates volumetric sampling in CT systems incorporating sparsely sampled detectors with axial and helical scan orbits and evaluates performance of model-based image reconstruction (MBIR) with spatially varying regularization in mitigating artifacts due to sparse detector sampling. Volumetric metrics of sampling density and uniformity were introduced. Penalized-likelihood MBIR with a spatially varying penalty that homogenized resolution by accounting for variations in local sampling density (i.e. detector gaps) was evaluated. The proposed methodology was tested in simulations and on an imaging bench based on a Si-strip PCD (total area 5 cm  ×  25 cm) consisting of an arrangement of line sensors separated by gaps of up to 2.5 mm. The bench was equipped with translation/rotation stages allowing a variety of scanning trajectories, ranging from a simple axial acquisition to helical scans with variable pitch. Statistical (spherical clutter) and anthropomorphic (hand) phantoms were considered. Image quality was compared to that obtained with a conventional uniform penalty in terms of structural similarity index (SSIM), image uniformity, spatial resolution, contrast, and noise. Scan trajectories with intermediate helical width (~10 mm longitudinal distance per 360° rotation) demonstrated optimal tradeoff between the average sampling density and the homogeneity of sampling throughout the volume. For a scan trajectory with 10.8 mm helical width, the spatially varying penalty resulted in significant visual reduction of sampling artifacts, confirmed by a 10% reduction in minimum SSIM (from 0.88 to 0.8) and a 40% reduction in the dispersion of SSIM in the volume compared to the constant penalty (both penalties applied at optimal regularization strength). Images of the spherical clutter and wrist phantoms confirmed the advantages of the spatially varying penalty, showing a 25% improvement in image uniformity and 1.8  ×  higher CNR (at matched spatial resolution) compared to the constant penalty. The studies elucidate the relationship between sampling in the detector plane, acquisition orbit, sampling of the reconstructed volume, and the resulting image quality. They also demonstrate the benefit of spatially varying regularization in MBIR for scenarios with irregular sampling patterns. Such findings are important and integral to the incorporation of a sparsely sampled Si-strip PCD in CT imaging.

Actively triggered 4d cone-beam CT acquisition

PURPOSE

4d cone-beam computed tomography (CBCT) scans are usually reconstructed by extracting the motion information from the 2d projections or an external surrogate signal, and binning the individual projections into multiple respiratory phases. In this "after-the-fact" binning approach, however, projections are unevenly distributed over respiratory phases resulting in inefficient utilization of imaging dose. To avoid excess dose in certain respiratory phases, and poor image quality due to a lack of projections in others, the authors have developed a novel 4d CBCT acquisition framework which actively triggers 2d projections based on the forward-predicted position of the tumor.

METHODS

The forward-prediction of the tumor position was independently established using either (i) an electromagnetic (EM) tracking system based on implanted EM-transponders which act as a surrogate for the tumor position, or (ii) an external motion sensor measuring the chest-wall displacement and correlating this external motion to the phase-shifted diaphragm motion derived from the acquired images. In order to avoid EM-induced artifacts in the imaging detector, the authors devised a simple but effective "Faraday" shielding cage. The authors demonstrated the feasibility of their acquisition strategy by scanning an anthropomorphic lung phantom moving on 1d or 2d sinusoidal trajectories.

RESULTS

With both tumor position devices, the authors were able to acquire 4d CBCTs free of motion blurring. For scans based on the EM tracking system, reconstruction artifacts stemming from the presence of the EM-array and the EM-transponders were greatly reduced using newly developed correction algorithms. By tuning the imaging frequency independently for each respiratory phase prior to acquisition, it was possible to harmonize the number of projections over respiratory phases. Depending on the breathing period (3.5 or 5 s) and the gantry rotation time (4 or 5 min), between ∼90 and 145 projections were acquired per respiratory phase resulting in a dose of ∼1.7-2.6 mGy per respiratory phase. Further dose savings and decreases in the scanning time are possible by acquiring only a subset of all respiratory phases, for example, peak-exhale and peak-inhale only scans.

CONCLUSIONS

This study is the first experimental demonstration of a new 4d CBCT acquisition paradigm in which imaging dose is efficiently utilized by actively triggering only those projections that are desired for the reconstruction process.

High-dose MVCT image guidance for stereotactic body radiation therapy

PURPOSE

Stereotactic body radiation therapy (SBRT) is a potent treatment for early stage primary and limited metastatic disease. Accurate tumor localization is essential to administer SBRT safely and effectively. Tomotherapy combines helical IMRT with onboard megavoltage CT (MVCT) imaging and is well suited for SBRT; however, MVCT results in reduced soft tissue contrast and increased image noise compared with kilovoltage CT. The goal of this work was to investigate the use of increased imaging doses on a clinical tomotherapy machine to improve image quality for SBRT image guidance.

METHODS

Two nonstandard, high-dose imaging modes were created on a tomotherapy machine by increasing the linear accelerator (LINAC) pulse rate from the nominal setting of 80 Hz, to 160 Hz and 300 Hz, respectively. Weighted CT dose indexes (wCTDIs) were measured for the standard, medium, and high-dose modes in a 30 cm solid water phantom using a calibrated A1SL ion chamber. Image quality was assessed from scans of a customized image quality phantom. Metrics evaluated include: contrast-to-noise ratios (CNRs), high-contrast spatial resolution, image uniformity, and percent image noise. In addition, two patients receiving SBRT were localized using high-dose MVCT scans. Raw detector data collected after each scan were used to reconstruct standard-dose images for comparison.

RESULTS

MVCT scans acquired using a pitch of 1.0 resulted in wCTDI values of 2.2, 4.7, and 8.5 cGy for the standard, medium, and high-dose modes respectively. CNR values for both low and high-contrast materials were found to increase with the square root of dose. Axial high-contrast spatial resolution was comparable for all imaging modes at 0.5 lp∕mm. Image uniformity was improved and percent noise decreased as the imaging dose increased. Similar improvements in image quality were observed in patient images, with decreases in image noise being the most notable.

CONCLUSIONS

High-dose imaging modes are made possible on a clinical tomotherapy machine by increasing the LINAC pulse rate. Increasing the imaging dose results in increased CNRs; making it easier to distinguish the boundaries of low contrast objects. The imaging dose levels observed in this work are considered acceptable at our institution for SBRT treatments delivered in 3-5 fractions.

Abnormal pixel detection using sum-of-projections symmetry in cone beam computed tomography

A novel abnormal pixels (APs) detection approach is proposed to remove artefacts from reconstructed images in cone beam computed tomography (CBCT). This approach is based on the symmetry detection of sum-of-projections (SOP). Because some factors affect the SOP symmetry, we combine dyadic wavelet transform-based singularity detection to extract the APs. Next, the Laplacian solution (LS) method is employed to restore the APs in each projection image. Experimental results demonstrate the efficiency of our approach for different imaging tasks.