Implementation of a channelized Hotelling observer model to assess image quality of x-ray angiography systems

Adaptation of a channelized Hotelling observer model to accommodate anthropomorphic backgrounds and moving test objects in X-ray angiography

BACKGROUND

Prior implementations of the channelized Hotelling observer (CHO) model have succeeded in assessing the performance of X-ray angiography systems under a variety of imaging conditions. However, often times these conditions do not resemble those present in routine clinical imaging scenarios, such as having complex anthropomorphic backgrounds in conjunction with moving test objects.

PURPOSE

This work builds up on prior established CHO methods and introduces a new approach to switch from the already established "multiple-sample" CHO implementation to a "single-sample" technique. The proposed implementation enables the inclusion of moving test objects upon nonuniform backgrounds by allowing only a single sample to represent the test object present condition that is to be used within the statistical test to estimate the detectability index.

METHODS

To assess the proposed method, two image data sets were acquired with a clinical X-ray angiography system. The first set consisted of a uniform background in combination with static test objects while the second consisted of an anthropomorphic chest phantom in conjunction with moving test objects. The first set was used to validate the proposed approach against the multiple-sample method while the second was used to assess the feasibility of the proposed method under a variety of imaging conditions, including seven object sizes and seven detector target dose (DTD) levels.

RESULTS

For the uniform background data set, considering all detectability indices greater or equal than 1, the ratio between the detectability indices of the proposed single-sample approach versus the multiple-sample method was 0.997 ± 0.056 (range 0.884-1.159). The average single-direction width of the 95% confidence intervals (CIs) of the detectability index estimates for the multiple-sample method was 0.38 ± 0.43 (range 0.03-2.20). For the single-sample approach, the average width was 2.52 ± 0.63 (range 1.11-5.44). For the anthropomorphic background image set, the results were consistent with classical quantum-limited signal-to-noise ratio (SNR) theory. The magnitude of the detectability indices varied predictably with changes in both object size and DTD, with the highest value associated with the highest dose and the largest object size. Additionally, the proposed method was able to capture differences in the imaging performance for a given test object across the field of view, which was associated with the attenuation levels provided by different features of the anthropomorphic background.

CONCLUSIONS

A new single-sample variant of the CHO model to assess the performance of X-ray angiography imaging systems is proposed. The new approach is consistent with quantum-limited image quality theory and with a standard implementation of the CHO model. The proposed method enables the assessment of moving test objects in combination with complex, nonuniform image backgrounds, thereby opening the possibility to assess imaging conditions which more closely resemble those used in clinical care.

Reducing the risk of hallucinations with interpretable deep learning models for low-dose CT denoising: comparative performance analysis

Objective.Reducing CT radiation dose is an often proposed measure to enhance patient safety, which, however results in increased image noise, translating into degradation of clinical image quality. Several deep learning methods have been proposed for low-dose CT (LDCT) denoising. The high risks posed by possible hallucinations in clinical images necessitate methods which aid the interpretation of deep learning networks. In this study, we aim to use qualitative reader studies and quantitative radiomics studies to assess the perceived quality, signal preservation and statistical feature preservation of LDCT volumes denoised by deep learning. We aim to compare interpretable deep learning methods with classical deep neural networks in clinical denoising performance.Approach.We conducted an image quality analysis study to assess the image quality of the denoised volumes based on four criteria to assess the perceived image quality. We subsequently conduct a lesion detection/segmentation study to assess the impact of denoising on signal detectability. Finally, a radiomic analysis study was performed to observe the quantitative and statistical similarity of the denoised images to standard dose CT (SDCT) images.Main results.The use of specific deep learning based algorithms generate denoised volumes which are qualitatively inferior to SDCT volumes(p< 0.05). Contrary to previous literature, denoising the volumes did not reduce the accuracy of the segmentation (p> 0.05). The denoised volumes, in most cases, generated radiomics features which were statistically similar to those generated from SDCT volumes (p> 0.05).Significance.Our results show that the denoised volumes have a lower perceived quality than SDCT volumes. Noise and denoising do not significantly affect detectability of the abdominal lesions. Denoised volumes also contain statistically identical features to SDCT volumes.

Experimental optimization of single-exposure dual-energy angiography with photon-counting x-ray detectors

BACKGROUND

Photon-counting x-ray detectors may enable single-exposure dual-energy (DE) x-ray angiography.

PURPOSE

The purpose of this paper is to experimentally optimize the energy thresholds and tube voltage for single-exposure DE x-ray angiography.

METHODS

We optimized single-exposure DE x-ray angiography using the iodine signal-difference-to-noise ratio (SDNR) per root patient air kerma (κ) as a figure of merit. We measured the iodine SDNR by imaging an iodine stepwedge immersed in a water tank with a depth of 30 cm in the direction of x-ray propagation. The stepwedge was imaged using tube voltages ranging from 90 to 150 kV and a cadmium telluride (CdTe) x-ray detector with two energy bins and analog charge summing for charge sharing suppression. The energy threshold that separates the two energy bins was varied from approximately 35 keV to approximately 75% of the maximum energy of the x-ray beam. Curve fitting was used to determine the threshold that maximized SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ . The effect of scatter was determined from measurements of the scatter-to-primary ratios (SPRs) of the low-energy and high-energy images and a semi-empirical model of the relationship between SDNR and SPR. Using the optimal parameters, we imaged a phantom with vessel-simulating structures and background clutter.

RESULTS

The optimal energy thresholds increased monotonically from ∼50 to ∼85 keV over the range of tube voltages considered. For tube voltages greater than 90 kV, the optimal energy thresholds consistently allocated approximately two thirds of all detected primary photons to the low energy bin; this ratio was preserved without scatter. Consistent with prior modeling studies, SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ increased monotonically with tube voltage from 90 to 150 kV; SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ at 150 kV was approximately 38% higher than that at 90 kV for an iodine area density of ∼50 mg/cm² . Scatter reduced SDNR by approximately 25% for SPRs of ∼1 and 0.4 in low-energy and high-energy images, respectively.

CONCLUSIONS

Achieving optimal image quality in single-exposure DE angiography with photon-counting x-ray detectors will require high tube voltages (i.e., >130 kV) and, for thick patients, energy thresholds that allocate approximately two thirds of all primary photons to the low-energy image. Future work will compare the image quality of singe-exposure photon-counting and kV-switching approaches to DE x-ray angiography.

Visual grading experiments and optimization in CBCT dental implantology imaging: preliminary application of integrated visual grading regression

This study uses a general formulation of integrated visual grading regression (IVGR) and applies it to cone beam computed tomography (CBCT) scan data related to anatomical landmarks for dental implantology. The aim was to assess and predict a minimum acceptable dose for diagnostic imaging and reporting. A skull phantom was imaged with a CBCT unit at various diagnostic exposures. Key anatomical landmarks within the images were independently reviewed by three trained observers. Each provided an overall image quality score. Statistical analysis was carried out to examine the acceptability of the images taken, using an IVGR analysis that was formulized as a three-stage protocol including defining an integrated score, development of an ordinal regression, and investigation of the possibility for dose reduction through estimated parameters. For a unit increase in the logarithm of radiation dose, the odds ratio that the integrated score for an image assessed by observers being rated in a higher category was 3.940 (95% confidence interval: 1.016-15.280). When assessed by the observers, the minimum dose required to achieve a 75% probability for an image to be classified as at least acceptable was 1346.91 mGy·cm² dose area product (DAP), a 31% reduction compared to the 1962 mGy·cm² DAP default dosage of the CBCT unit. The kappa values of the intra and inter-observer reliability indicated moderate agreements, while a discrepancy among observers was also identified because each, as expected, perceived visibility differently. The results of this work demonstrate the IVGR's predictive value of dose saving in the effort to reduce dose to patients while maintaining reportable diagnostic image quality.

Medical image quality metrics for foveated model observers

Purpose: A recently proposed model observer mimics the foveated nature of the human visual system by processing the entire image with varying spatial detail, executing eye movements, and scrolling through slices. The model can predict how human search performance changes with signal type and modality (2D versus 3D), yet its implementation is computationally expensive and time-consuming. Here, we evaluate various image quality metrics using extensions of the classic index of detectability expression and assess foveated model observers for search tasks. Approach: We evaluated foveated extensions of a channelized Hotelling and nonprewhitening matched filter model with an eye filter. The proposed methods involve calculating a model index of detectability ( d ' ) for each retinal eccentricity and combining these with a weighting function into a single detectability metric. We assessed different versions of the weighting function that varied in the required measurements of the human observers' search (no measurements, eye movement patterns, size of the image, and median search times). Results: We show that the index of detectability across eccentricities weighted using the eye movement patterns of observers best predicted human performance in 2D versus 3D search performance for a small microcalcification-like signal and a larger mass-like. The metric with a weighting function based on median search times was the second best predicting human results. Conclusions: The findings provide a set of model observer tools to evaluate image quality in the early stages of imaging system evaluation or design without implementing the more computationally complex foveated search model.

Task-specific efficient channel selection and bias management for Gabor function channelized Hotelling observer model for the assessment of x-ray angiography system performance

PURPOSE

Channelized Hotelling observer (CHO) models have been implemented to assess imaging performance in x-ray angiography systems. While current methods are appropriate for assessing unprocessed images of moving test objects upon uniform-exposure backgrounds, they are inadequate for assessing conditions which more appropriately mimic clinical imaging conditions including the combination of moving test objects, complex anthropomorphic backgrounds, and image processing. In support of this broad goal, the purpose of this work was to develop theory and methods to automatically select a subset of task-specific efficient Gabor channels from a task-generic Gabor channel base set. Also, previously described theory and methods to manage detectability index (d') bias due to nonrandom temporal variations in image electronic noise will be revisited herein.

METHODS

Starting with a base set of 96 Gabor channels, backward elimination of channels was used to automatically identify an "efficient" channel subset which reduced the number of channels retained in the subset while maintaining the magnitude of the d' estimate. The concept of a pixelwise Hotelling observer (PHO) model was introduced and similarly implemented to assess the performance of the efficient-channel CHO model. Bias in d' estimates arising from temporally variable nonstationary noise was modeled as a bivariate probability density function for normal distributions, where one variable corresponds to the signal from the test object and the other variable corresponds to the signal from temporally variable nonstationary noise. Theory and methods were tested on uniform-exposure unprocessed angiography images with detector target dose (DTD) of 6, 18, and 120 nGy containing static disk-shaped test objects with diameter in the range of 0.5 to 4 mm.

RESULTS

Considering all DTD levels and test object sizes, the proposed method reduced the number of Gabor channels in the final subset by 63-82% compared to the original 96 Gabor channel base set, while maintaining a mean relative performance ( ( d CHO ' / d PHO ' ) × 100 % ) of 95%  ±  4% with respect to the reference PHO model. Experimental results demonstrated that the bivariate approach to account for bias due to temporally variable nonstationary noise resulted in improved correlation between the CHO and PHO models as compared to a previously proposed univariate approach.

CONCLUSIONS

Computationally efficient backward elimination can be used to select an efficient subset of Gabor channels from an initial channel base set without substantially compromising the magnitude of the d' estimate. Bias due to temporally variable nonstationary noise can be modeled through a bivariate approach leading to an improved unbiased estimate of d'.

Foveated Model Observers for Visual Search in 3D Medical Images

Model observers have a long history of success in predicting human observer performance in clinically-relevant detection tasks. New 3D image modalities provide more signal information but vastly increase the search space to be scrutinized. Here, we compared standard linear model observers (ideal observers, non-pre-whitening matched filter with eye filter, and various versions of Channelized Hotelling models) to human performance searching in 3D 1/f2.8 filtered noise images and assessed its relationship to the more traditional location known exactly detection tasks and 2D search. We investigated two different signal types that vary in their detectability away from the point of fixation (visual periphery). We show that the influence of 3D search on human performance interacts with the signal's detectability in the visual periphery. Detection performance for signals difficult to detect in the visual periphery deteriorates greatly in 3D search but not in 3D location known exactly and 2D search. Standard model observers do not predict the interaction between 3D search and signal type. A proposed extension of the Channelized Hotelling model (foveated search model) that processes the image with reduced spatial detail away from the point of fixation, explores the image through eye movements, and scrolls across slices can successfully predict the interaction observed in humans and also the types of errors in 3D search. Together, the findings highlight the need for foveated model observers for image quality evaluation with 3D search.

Radiation dose reduction and static image quality assessment using a channelized hotelling observer on an angiography system upgraded with clarity IQ

The goal of this paper was the comparison of radiation dose and imaging quality before and after the Clarity IQ technology installation in a Philips AlluraXper FD20/20 angiography system using a Channelized Hotelling Observer model (CHO). The core characteristics of the Allura Clarity IQ technology are its real-time noise reduction algorithms (NRT) combined with state-of-the-art hardware; this technology allows to implement acquisition protocols able to significantly reduce patient entrance dose. To measure the system performances in terms of image quality we used a contrast detail phantom in a clinical scatter condition. A Leeds TO10 phantom has been imaged between two 10 cm thick homogeneous solid water slabs. Fluoroscopy images were acquired using a cerebral protocol at 3 dose levels (low, medium and high) with a field- of view (FOV) of 31 cm. Cineangiography images were acquired using a cerebral protocol at 2 fps. Thus, 4 acquisitions were obtained for the conventional technology and 4 acquisitions were taken after the Clarity IQ upgrade, for a total of 8 different image sets. A validated 40 Gabor channels CHO with an internal noise model compared the image sets. Human observers' studies were carried out to tune the internal noise parameter. We showed that the CHO did not detect any significant difference between any of the image sets acquired using the two technologies. Consequently, this x-ray imaging technology provides a non-inferior image quality with an average patient dose reduction of 57% and 28% respectively in cineangiography and fluoroscopy. The Clarity IQ installation has certainly allowed a considerable improvement in patient and staff safety, while maintaining the same image quality.

Evaluation of Convolutional Neural Networks for Search in 1/f 2.8 Filtered Noise and Digital Breast Tomosynthesis Phantoms

With the advent of powerful convolutional neural networks (CNNs), recent studies have extended early applications of neural networks to imaging tasks thus making CNNs a potential new tool for assessing medical image quality. Here, we compare a CNN to model observers in a search task for two possible signals (a simulated mass and a smaller simulated micro-calcification) embedded in filtered noise and single slices of Digital Breast Tomosynthesis (DBT) virtual phantoms. For the case of the filtered noise, we show how a CNN can approximate the ideal observer for a search task, achieving a statistical efficiency of 0.77 for the microcalcification and 0.78 for the mass. For search in single slices of DBT phantoms, we show that a Channelized Hotelling Observer (CHO) performance is affected detrimentally by false positives related to anatomic variations and results in detection accuracy below human observer performance. In contrast, the CNN learns to identify and discount the backgrounds, and achieves performance comparable to that of human observer and superior to model observers (Proportion Correct for the microcalcification: CNN = 0.96; Humans = 0.98; CHO = 0.84; Proportion Correct for the mass: CNN = 0.98; Humans = 0.83; CHO = 0.51). Together, our results provide an important evaluation of CNN methods by benchmarking their performance against human and model observers in complex search tasks.

Determination of Optimal Image Type and Lowest Detectable Concentration for Iodine Detection on a Photon Counting Detector-Based Multi-Energy CT System

Photon counting detector (PCD) based multi-energy CT is able to generate different types of images such as virtual monoenergetic images (VMIs) and material specific images (e.g., iodine maps) in addition to the conventional single energy images. The purpose of this study is to determine the image type that has optimal iodine detection and to determine the lowest detectable iodine concentration using a PCD-CT system. A 35 cm body phantom with iodine inserts of 4 concentrations and 2 sizes was scanned on a research PCD-CT system. For each iodine concentration, 80 repeated scans were performed and images were reconstructed for each energy threshold. In addition, VMIs at different keVs and iodine maps were also generated. CNR was measured for each type of images. A channelized Hotelling observer was used to assess iodine detectability after being validated with human observer studies, with area under the ROC curve (AUC) as a figure of merit. The agreement between model and human observer performance indicated that model observer could serve as an effective approach to determine optimal image type for the clinical practice and to determine the lowest detectable iodine concentration. Results demonstrated that for all size and concentration combinations, VMI at 70 keV had similar performance as that of threshold low images, both of which outperformed the iodine map images. At the AUC value of 0.8, iodine concentration as low as 0.2 mgI/cc could be detected for an 8 mm object and 0.5 mgI/cc for a 4 mm object with a 5 mm slice thickness.

Bayesian framework inspired no-reference region-of-interest quality measure for brain MRI images

We describe a postacquisition, attribute-based quality assessment method for brain magnetic resonance imaging (MRI) images. It is based on the application of Bayes theory to the relationship between entropy and image quality attributes. The entropy feature image of a slice is segmented into low- and high-entropy regions. For each entropy region, there are three separate observations of contrast, standard deviation, and sharpness quality attributes. A quality index for a quality attribute is the posterior probability of an entropy region given any corresponding region in a feature image where quality attribute is observed. Prior belief in each entropy region is determined from normalized total clique potential (TCP) energy of the slice. For TCP below the predefined threshold, the prior probability for a region is determined by deviation of its percentage composition in the slice from a standard normal distribution built from 250 MRI volume data provided by Alzheimer's Disease Neuroimaging Initiative. For TCP above the threshold, the prior is computed using a mathematical model that describes the TCP-noise level relationship in brain MRI images. Our proposed method assesses the image quality of each entropy region and the global image. Experimental results demonstrate good correlation with subjective opinions of radiologists for different types and levels of quality distortions.

Direct estimation and correction of bias from temporally variable non-stationary noise in a channelized Hotelling model observer

Channelized Hotelling model observer (CHO) methods were developed to assess performance of an x-ray angiography system. The analytical methods included correction for known bias error due to finite sampling. Detectability indices ([Formula: see text]) corresponding to disk-shaped objects with diameters in the range 0.5-4 mm were calculated. Application of the CHO for variable detector target dose (DTD) in the range 6-240 nGy frame(-1) resulted in [Formula: see text] estimates which were as much as 2.9×  greater than expected of a quantum limited system. Over-estimation of [Formula: see text] was presumed to be a result of bias error due to temporally variable non-stationary noise. Statistical theory which allows for independent contributions of 'signal' from a test object (o) and temporally variable non-stationary noise (ns) was developed. The theory demonstrates that the biased [Formula: see text] is the sum of the detectability indices associated with the test object [Formula: see text] and non-stationary noise ([Formula: see text]). Given the nature of the imaging system and the experimental methods, [Formula: see text] cannot be directly determined independent of [Formula: see text]. However, methods to estimate [Formula: see text] independent of [Formula: see text] were developed. In accordance with the theory, [Formula: see text] was subtracted from experimental estimates of [Formula: see text], providing an unbiased estimate of [Formula: see text]. Estimates of [Formula: see text] exhibited trends consistent with expectations of an angiography system that is quantum limited for high DTD and compromised by detector electronic readout noise for low DTD conditions. Results suggest that these methods provide [Formula: see text] estimates which are accurate and precise for [Formula: see text]. Further, results demonstrated that the source of bias was detector electronic readout noise. In summary, this work presents theory and methods to test for the presence of bias in Hotelling model observers due to temporally variable non-stationary noise and correct this bias when the temporally variable non-stationary noise is independent and additive with respect to the test object signal.

SYSTEM UPGRADE ON PHILIPS ALLURA FD20 ANGIOGRAPHY SYSTEMS: EFFECTS ON PATIENT SKIN DOSE AND STATIC IMAGE QUALITY

Fluoroscopically guided procedures might be highly irradiating for patients, possibly leading to skin injuries. In such a context, every effort should be done to lower patient exposure as much as possible. Moreover, patient dose reduction does not only benefit to the patient but also allows reducing staff exposure. In this framework, Philips Healthcare recently introduced a system upgrade for their angiography units, called 'AlluraClarity'. The authors performed air kerma rate measurements for all available fluoroscopy modes and air kerma per frame measurements for the digital subtraction angiography protocols, along with subjective spatial resolution and low-contrast detectability assessments using a standard QA phantom. Air kerma reductions ranging from 25.5 to 84.4 % were found, with no significant change in image quality when switching from a standard operating mode to an upgraded version. These results are confirmed by the comparison of actual patient exposures for similar procedures.