CT image quality metrics implementation during thorax SPECT/CT clinic studies

Image quality metrics have become more and more popular in the image processing community. However, so far, no one has been able to define an image quality metric well correlated with the percept for overall image quality. One of the causes is that image quality is multi-dimensional and complex. One approach to bridge the gap between perceived and calculated image quality is to reduce the complexity of image quality, by breaking the overall quality into a set of quality attributes. In our research we have presented a set of quality attributes built on existing attributes from the literature. The six proposed quality attributes are: sharpness, color, lightness, artifacts, contrast, and physical. This set keeps the dimensionality to a minimum. An experiment validated the quality attributes as suitable for image quality evaluation. The process of applying image quality metrics to printed images is not straightforward, because image quality metrics require a digital input. A framework has been developed for this process, which includes scanning the print to get a digital copy, image registration, and the application of image quality metrics. With quality attributes for the evaluation of image quality and a framework for applying image quality metrics, a selection of suitable image quality metrics for the different quality attributes has been carried out. Each of the quality attributes has been investigated, and an experimental analysis carried out to find the most suitable image quality metrics for the given quality attributes. For the sharpness attributes the Structural SIMilarity index (SSIM) by Wang et al. (2004) is the the most suitable, and for the other attributes further evaluation is required.

Motion-induced artifact detection has become a fixture in the assessment of functional magnetic resonance imaging (fMRI) quality control. However, the effects of other MR image quality (IQ) metrics on intrinsic connectivity brain networks are largely unexplored. Accordingly, we report herein the initial assessment of the effects of a comprehensive list of IQ metrics on resting state networks using a multivariate analysis of covariance (MANCOVA) approach based on high-order spatial independent component analysis (ICA). Three categories of MR IQ metrics were considered: (1) metrics for artifacts including the AFNI outlier ratio and quality index, framewise displacement, and ghost to signal ratio, (2) metrics for the temporal quality of MRI data including the temporal framewise change in global BOLD signals (DVARS), global correlation of time-series, and temporal signal to noise ratio, (3) metrics for the structural quality of MRI data including the entropy focus criterion, foreground-background energy ratio, full-width half maximum smoothness, and static signal to noise ratio. After FDR-correction for multiple comparisons, results showed significant effects of the static and temporal signal to noise ratios on the spatial map intensities of the basal ganglia, default-mode and cerebellar networks. AFNI outlier ratio, framewise displacement and DVARS exhibited significant effects on the BOLD power spectra of sensorimotor networks. The global correlation of time-series displayed wide-spread modulation of the spectral power in most networks. Further investigations of the effect of IQ metrics on the characteristics of intrinsic connectivity brain networks allow more accurate interpretation of the fMRI results.

© IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

For practicing medical physicists, CT image quality assessment is, in general, limited to the ones required by ACR and based on the axial images, such as CT number uniformity, image noise, contrast‐to‐noise ratio (CNR), and high contrast resolution. However, with more advanced technologies introduced by CT vendors, just assessing these basic image quality metrics may not be enough. Noise power spectrum (NPS), characterizing the signal noise component at each spatial frequency, is a powerful tool to predict the CT performance related to lesion detectability and noise texture. The quality of coronal and sagittal reformatted images is, in a large part, determined by the slice profile. How to accurately assess the slice profile is important for clinical medical physicists but not well recognized.Automatic tube current modulation (ATCM) has been adopted by major CT manufacturers, to maintain image quality across the scan range. The way how ATMC works varies significantly among vendors. A “standard” method is needed to evaluate the performance of this critical dose‐saving and image quality improvement technique. As dual energy CT becomes more readily available for routine clinical use, the need of image quality assessment metrics unique to the dual energy CT mode is becoming urgent.The session will facilitate the discussion of NPS and its application in the evaluation of FBP and iterative reconstruction. This session will also present the methods to measure the slice profile and characterize the performance of ATCM. The last part of this session will focus on the physics of CT dual energy mode, as well as unique image quality metrics to evaluate the performance of the dual energy mode.Learning Objectives: Be familiar with procedures on how to correctly compute CT noise power spectrum Understand why basic image quality metrics used for FBP may not be sufficient to characterize the performance of advanced iterative reconstruction Understand CT slice profile and how to accurately measure it Be familiar with ATCM from major CT vendors and how to evaluate its performance Understand different hardware approaches for dual energy CT Be familiar with image quality metrics relevant to dual energy mode Schmidt: Employee of Siemens HealthcareFan: Employee of GE Healthcare

Image quality metrics have been widely used in imaging systems to maintain and improve the quality of images being processed and transmitted. Due to the close relationship between image quality perception and the human visual system, the development of image quality metrics has been contributed to by both psychologists and computer scientists. In this paper, three novel image quality metrics are proposed by improving the well-known image quality metric structural similarity index (SSIM). In this new approach, images are not compared directly, but their feature maps are (preprocessing is incorporated to extract the corner, edge and symmetry maps). The similarity measured (by SSIM) between corner, edge and symmetry maps of images being compared is used as an indicator of image quality, and named C SSIM, E SSIM and S SSIM respectively. The experiments show that all the proposed image quality metrics have a better performance than SSIM, and E SSIM has the best performance among them.

When an investigator is developing a new image-processing technique or manipulating an image-acquisition technique they are confronted with the question of whether the new technique will improve clinical diagnosis. A first approach is to look at individual physical properties of the image such as image contrast and resolution. Although these properties might be useful, it has long been known that the noise characteristics of the image system need to be taken into consideration to appropriately evaluate the quality of an image whether it will be used to detect, classify, and/ˆ•or estimate a signal (Cunningham and Shaw, 1999). One useful measure of the noise characteristics is the noise-equivalent quanta (NEQ) that expresses the image noise in terms of the number of Poisson-distributed input photons per unit area at each spatial frequency (Wagner and Brown, 1985). The NEQ can be thought of as a measure inversely related to the amount of noise as a function of spatial frequency. However, when the diagnostic decision involves a human observer, medical image quality can be defined in terms of human performance in visual tasks that are relevant to clinical diagnosis (Barrett, 1993). In this context, receiver operating characteristics (ROC) studies are the standard method of evaluating the impact of a particular image manipulation on clinical diagnosis. In these studies, the physicians scrutinize a set of medical images (under the different image-acquisition or processing conditions) and rate their confidence about the presence of the lesion. The investigator infers from these ratings a measure of performance known as the area under the ROC curve. Often, the number of possible conditions is large and ROC studies become time consuming and costly because they require a large number of human observations. Other times, the investigator might want to optimize a parameter or a set of parameters. In such cases, the number of conditions suffers a combinatorial explosion, and therefore ROC studies become unfeasible. Thus it is desirable to develop a metric of image quality that could be used for fast evaluation and optimization of image quality but also would have the predictive power of ROC studies. Computer-model observers are algorithms that attempt to predict human visual performance in noisy images and might represent the desired metric of image quality when the diagnostic decision involves a human observer and a visual task. Development of models to predict human visual signal detection in noise goes back to work by Rose (1948) who studied the detectability of a flat-topped disk embedded in white noise (see Burgess, 1999a, for a review). In the last two decades, many studies have concentrated on finding a model observer that can predict human performance across many types of synthetic backgrounds. More recently, model observers have been applied to real medical-image backgrounds. The hope is that eventually model observers will become common metrics of task-based image quality for evaluation of medical-image quality as well as optimization of imaging systems.

To evaluate the effects of arm positioning and reconstruction algorithms on radiation dose and image quality of abdominal CT. Abdominal CT scans were performed using a customized body phantom at 100 kVp with automatic tube current modulation based on noise indices (NI) of 9, 11, and 13 across seven arm positions: arms up (AU), arms down (AD), arms down with one or two-layered air cushions (ADAC, ADAC2), arms on the belly (AB), and arms on the belly with one or two-layered air cushions (ABAC, ABAC2). Images were reconstructed using filtered back projection (FBP), iterative reconstruction, and deep learning-based reconstruction (DLIR). Radiation dose was recorded. Quantitative image quality metrics included image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), sharpness, and structural similarity. Qualitative assessment of sharpness, noise, artifacts, and overall image quality was performed using fixed-dose AU-FBP images as the reference standard. Compared with the AU position, radiation dose increased by up to 83% in non-AU configurations, with similar trends across all NI settings. In contrast, air cushion use caused only minor, configuration-dependent dose changes. Qualitatively, non-AU FBP images showed marked degradation, whereas DLIR enabled selected non-AU configurations to approach near-reference diagnostic quality. Air cushion-related benefits were modest overall but more apparent in AB than AD configurations. Arm positioning is the primary determinant of radiation dose and image quality in abdominal CT. When arm elevation is not feasible, DLIR effectively mitigates image quality degradation, while air cushions provide limited, configuration-dependent benefits. When arms cannot be raised during abdominal CT, avoiding arms-on-belly and using air cushions may help maintain image quality without substantially increasing radiation dose, while deep learning image reconstruction further enhances diagnostic performance under restricted positioning conditions. Arm positioning is a major determinant of radiation dose and image quality in abdominal CT. Arms-on-belly positioning produces poorer image quality despite increased radiation dose. DLIR improves image quality across arm positions but does not eliminate positioning-related degradation.

This study aimed to evaluate the feasibility and accuracy of automated contrast-to-noise ratio (CNR) analysis in chest CT using the open-source body and organ analysis (BOA) framework and to validate segmentation modifications for reproducible image-quality assessment. This retrospective study analyzed 100 contrast-enhanced chest CTs (mean age 60.2 ± 15 years; 40% female; 50 CTA, 50 CTPA) and validated the approach in an external cancer imaging archive (TCIA) cohort (n = 100). Automated BOA segmentations of the aorta, pulmonary trunk, and paraspinal muscles were modified by fat subtraction and binary erosion and compared with manual measurements from three radiologists. Agreement was assessed using statistical testing, Bland-Altman analysis, and intraclass correlation coefficients (ICC). Unmodified BOA segmentations yielded significantly lower CNRs than manual measurements (all p < 0.01, mean difference up to 6.3). Fat subtraction and binary erosion progressively reduced deviations, with the optimized variant (m_erode6 combined with a_erode6 or p_erode6) showing no significant differences from radiologists (p ≥ 0.35). In the external TCIA validation cohort (n = 100), agreement was excellent (ICC 0.89-0.93), and Bland-Altman analysis demonstrated minimal bias (Aorta: 0.16 [limits of agreement (LoA) -3.47 to 3.80]; PT: 0.42 [LoA -4.03 to 4.87]). A minimally modified open-source segmentation framework enables fully automated, reproducible CNR assessment in chest CT, achieving expert-level agreement, including robust performance in external validation. This scalable alternative to manual region-of-interest (ROI) measurement streamlines image-quality assessment, facilitates protocol optimization, and provides standardized metrics ready for integration into AI workflows. This study provides a validated, fully automated method for quantitative CT image quality assessment, reducing observer dependence and enabling consistent evaluation across scanners, protocols, and institutions, thereby supporting reproducible image quality metrics in clinical routine. Automated CNR assessment enables objective and reproducible evaluation of image quality in CTA and CTPA. Adjustments of the segmentation strategy can substantially improve the accuracy of automated measurements. The fully automated approach provides a foundation for standardized and scalable CT image quality analysis in research and clinical practice.

To assess the performance of routine cone-beam computed tomography (CBCT) quality assurance (QA) at predicting and diagnosing clinically recognizable linac CBCT image quality issues. Monthly automated linac CBCT image quality QA data were acquired on eight Varian linacs (Varian Medical Systems, Palo Alto, CA) using the CATPHAN 500 series phantom (The Phantom Laboratory, Inc., Greenwich, NY) and Total QA software (Image Owl, Inc., Greenwich, NY) over 34months between July 2014 and May 2017. For each linac, the following image quality metrics were acquired: geometric distortion, spatial resolution, Hounsfield Unit (HU) constancy, uniformity, and noise. Quality control (QC) limits were determined by American Association of Physicists in Medicine (AAPM) expert consensus documents Task Group (TG-142 and TG-179) and the manufacturer acceptance testing procedure. Clinically recognizable CBCT issues were extracted from the in-house incident learning system (ILS) and service reports. The sensitivity and specificity of CATPHAN QA at predicting clinically recognizable image quality issues was investigated. Sensitivity was defined as the percentage of clinically recognizable CBCT image quality issues that followed a failing CATPHAN QA. Quality assurance results are categorized as failing if one or more image quality metrics are outside the QC limits. The specificity of CATPHAN QA was defined as one minus the fraction of failing CATPHAN QA results that did not have a clinically recognizable CBCT image quality issue in the subsequent month. Receiver operating characteristic (ROC) curves were generated for each image quality metric by plotting the true positive rate (TPR) against the false-positive rate (FPR). Over the study period, 18 image quality issues were discovered by clinicians while using CBCT to set up the patient and five were reported prior to x-ray tube repair. The incidents ranged from ring artifacts to uniformity problems. The sensitivity of the TG-142/179 limits was 17% (four of the prior monthly QC tests detected a clinically recognizable image quality issue). The area under the curve (AUC) calculated for each image quality metric ROC curve was: 0.85 for uniformity, 0.66 for spatial resolution, 0.51 for geometric distortion, 0.56 for noise, 0.73 for HU constancy, and 0.59 for contrast resolution. Automated monthly QA is not a good predictor of CBCT image quality issues. Of the available metrics, uniformity has the best predictive performance, but still has a high FPR and low sensitivity. The poor performance of CATPHAN QA as a predictor of image quality problems is partially due to its reliance on region-of-interest (ROI) based algorithms and a lack of a global algorithm such as correlation. The manner in which image quality issues occur (trending toward failure or random) is still not known and should be studied further. CBCT image quality QA should be adapted based on how CBCT is used clinically.

For image quality assessment, the availability of diverse databases is vital for the development and evaluation of image quality metrics. Existing databases have played an important role in promoting the understanding of various types of distortion and in the evaluation of image quality metrics. However, a comprehensive representation of optical aberrations and their impact on image quality is lacking. This paper addresses this gap by introducing a novel image quality database that focuses on optical aberrations. We conduct a subjective experiment to capture human perceptual responses on a set of images with optical aberrations. We then test the performance of selected objective image quality metrics to assess these aberrations. This approach not only ensures the relevance of our database to real-world scenarios but also contributes to ensuring the performance of the selected image quality metrics. The database is available for download at https: // www. ntnu. edu/ colourlab/ software .

This paper presents a new area-efficient approximate 4-2 compressor with improved image quality and error metrics. The proposed compressor is developed using functional approximation and Karnaugh map (K-map) simplification. The circuit behavior is verified through modeling and synthesis using Verilog HDL. The effectiveness of the proposed compressor for image multiplication is evaluated using an 8 × 8 approximate multiplier simulated in MATLAB. The performance of the proposed design is compared with existing compressors using circuit, error, and image-quality metrics. Post-synthesis and image-multiplication results show that the proposed compressor achieves excellent power, delay, and image-quality performance compared to existing area-efficient designs.

Ground-based solar observations are severely affected by Earth’s turbulent atmosphere. As a consequence, observed image quality and prevailing seeing conditions are closely related. Partial correction of image degradation is nowadays provided in real time by adaptive optics (AO) systems. In this study, different metrics of image quality are compared with parameters characterizing the prevailing seeing conditions, i.e. Median Filter Gradient Similarity (MFGS), Median Filter Laplacian Similarity (MFLS), Helmli–Scherer mean, granular rms-contrast, differential image motion, and Fried-parameter \(r_{0}\). The quiet-Sun observations at disk center were carried out at the Vacuum Tower Telescope (VTT), Observatorio del Teide (OT), Izaña, Tenerife, Spain. In July and August 2016, time series of short-exposure images were recorded with the High-resolution Fast Imager (HiFI) at various wavelengths in the visible and near-infrared parts of the spectrum. Correlation analysis yields the wavelength dependence of the image quality metrics and seeing parameters, and Uniform Manifold Approximation and Projection (UMAP) is employed to characterize the seeing on a particular observing day. In addition, the image quality metrics and seeing parameters are used to determine the field dependence of the correction provided by the AO system. Management of high-resolution imaging data from large-aperture, ground-based telescopes demands reliable image quality metrics and meaningful characterization of prevailing seeing conditions and AO performance. The present study offers guidance on how retrieving such information ex post facto.

To investigate the repeatability of a combined Dual-Scheimpflug placido disc corneal tomographer/topographer (Ziemer Galilei G4) with respect to keratometric indices used to monitor progression of keratoconus (KCN). Patients with KCN were prospectively enrolled. For each eye lacking history of corneal surgery, 5 measurements were taken in succession. Eyes in which 3 or more measurements could be obtained (defined by the device's 4 image quality metrics) were included in the analysis. The repeatability limits (RL) and interclass correlation coefficients (ICC) were calculated for various parameters. Thirty-two eyes from 25 patients met all image quality metrics, and 54 eyes from 38 patients met at least 3/4 criteria (all except the placido image quality metric). RLs for key parameters when 4/4 or ≥3/4 image quality metrics were met included: 0.37 and 0.77 diopters (D) for steep simulated keratometry, 0.79 and 1.65 D for maximum keratometry, 13.80 and 13.88 degrees for astigmatism axis, 0.64 and 0.56µm for vertical coma magnitude, and 3.76 and 3.84µm for thinnest pachymetry, respectively. The ICCs for all parameters were excellent (above 0.87) except for spherical aberration (0.77), which was still considered good. The dual-Scheimpflug placido disc corneal tomographer/topographer is highly repeatable in quantifying parameters used in monitoring KCN. Excellent placido images are difficult to capture in eyes with KCN, but when available, increase the reliability of the measurements. When clinicians find that a topographic index changes by more than the RLs defined herein, they can have confidence that this represents real change and may appropriately recommend interventions such as corneal cross-linking or intrastromal corneal ring segments.

Purpose: As large bore CT scanners are becoming more prevalent in the marketplace, so have questions about the image quality and dose associated with these scanners. The aim of this work is to compare the image quality and dose characteristics of a large bore 16‐slice CT scanner with a standard 16‐slice CT scanner. Methods: Two CT scanners, a large bore CT (Aquilion Large Bore, Toshiba) and a standard 16‐slice CT (Aquilion 16, Toshiba) were compared with respect to noise, spatial resolution, CT number accuracy, and low contrast. The technique factors on the large bore were adjusted to yield the same dose as a standard abdominal protocol on a standard 16‐slice scanner. Noise was measured on a 24cm water phantom, spatial resolution via the Fourier transform of bead, CT number via several known materials contained in a 32cm diameter phantom, and low contrast via a traditional low contrast phantom. Dose was measured via standard CTDI measurements. Results: The large bore required an increase in mAs of approximately 30% to yield the same dose as a standard 16‐slice scanner. After this adjustment was made, the image quality metrics for both scanners were identical. Conclusion: The large bore CT in this study yielded the same image quality in all four categories as the standard 16‐slice protocol.K. Boedeker, R. Mather, and M. MacLeod are employees of Toshiba America Medical Systems.

Magnetic resonance imaging (MRI) scans are known to suffer from a variety of acquisition artifacts as well as equipment-based variations that impact image appearance and segmentation performance. It is still unclear whether a direct relationship exists between magnetic resonance (MR) image quality metrics (IQMs) (e.g., signal-to-noise, contrast-to-noise) and segmentation accuracy. Deep learning (DL) approaches have shown significant promise for automated segmentation of brain tumors on MRI but depend on the quality of input training images. We sought to evaluate the relationship between IQMs of input training images and DL-based brain tumor segmentation accuracy toward developing more generalizable models for multi-institutional data. We trained a 3D DenseNet model on the BraTS 2020 cohorts for segmentation of tumor subregions enhancing tumor (ET), peritumoral edematous, and necrotic and non-ET on MRI; with performance quantified via a 5-fold cross-validated Dice coefficient. MRI scans were evaluated through the open-source quality control tool MRQy, to yield 13 IQMs per scan. The Pearson correlation coefficient was computed between whole tumor (WT) dice values and IQM measures in the training cohorts to identify quality measures most correlated with segmentation performance. Each selected IQM was used to group MRI scans as "better" quality (BQ) or "worse" quality (WQ), via relative thresholding. Segmentation performance was re-evaluated for the DenseNet model when (i) training on BQ MRI images with validation on WQ images, as well as (ii) training on WQ images, and validation on BQ images. Trends were further validated on independent test sets derived from the BraTS 2021 training cohorts. For this study, multimodal MRI scans from the BraTS 2020 training cohorts were used to train the segmentation model and validated on independent test sets derived from the BraTS 2021 cohort. Among the selected IQMs, models trained on BQ images based on inhomogeneity measurements (coefficient of variance, coefficient of joint variation, coefficient of variation of the foreground patch) and the models trained on WQ images based on noise measurement peak signal-to-noise ratio (SNR) yielded significantly improved tumor segmentation accuracy compared to their inverse models. Our results suggest that a significant correlation may exist between specific MR IQMs and DenseNet-based brain tumor segmentation performance. The selection of MRI scans for model training based on IQMs may yield more accurate and generalizable models in unseen validation.