X-ray imaging optimization using virtual phantoms and computerized observer modelling

This study develops and demonstrates a realistic x-ray imaging simulator with computerized observers to maximize lesion detectability and minimize patient exposure. A software package, ViPRIS, incorporating two computational patient phantoms, has been developed for simulating x-ray radiographic images. A tomographic phantom, VIP-Man, constructed from Visible Human anatomical colour images is used to simulate the scattered portion using the ESGnrc Monte Carlo code. The primary portion of an x-ray image is simulated using the projection ray-tracing method through the Visible Human CT data set. To produce a realistic image, the software simulates quantum noise, blurring effects, lesions, detector absorption efficiency and other imaging artefacts. The primary and scattered portions of an x-ray chest image are combined to form a final image for computerized observer studies and image quality analysis. Absorbed doses in organs and tissues of the segmented VIP-Man phantom were also obtained from the Monte Carlo simulations. Approximately 25,000 simulated images and 2,500,000 data files were analysed using computerized observers. Hotelling and Laguerre-Gauss Hotelling observers are used to perform various lesion detection tasks. Several model observer tasks were used including SKE/BKE, MAFC and SKEV. The energy levels and fluence at the minimum dose required to detect a small lesion were determined with respect to lesion size, location and system parameters.     

Signal detectability in digital radiography: spatial domain figures of merit

The usefulness of Fourier-based measures of imaging performance has come into question for the evaluation of digital imaging systems. Figures of merit such as detective quantum efficiency (DQE) based on Fourier domain parameters are relevant for linear, shift-invariant systems with stationary noise. However, no digital imaging system is shift invariant, and realistic images do not satisfy the stationarity condition. Our methods for the task-based evaluation of imaging systems, based on signal detectability in the spatial domain, do not require such assumptions. We have computed the performance of ideal and quasi-ideal observers for the task of signal detection in digital radiography. Signal detectability in terms of an observer signal-to-noise-ratio (SNR) has been compared to results obtained from a Monte Carlo simulation of the digital image-acquisition process. The simulation incorporates the effects of random amplification and secondary quantum blur, integration over pixel area, and electronic noise. The observer figures of merit that have been previously shown to bracket human performance directly specify the usefulness of the images for the stated diagnostic task. In addition, the observer figures of merit give a task-dependent measure of imaging system efficiency in terms of the ratio of an output SNR2 to an input SNR2. Thus, the concept of "detective quantum efficiency" reappears in a natural way but based in the spatial domain and not dependent on shift invariance and stationarity assumptions. With respect to the optimum amount of system blur, our simulations indicate that under certain task-dependent conditions, large signals are fairly insensitive to blur in the x-ray transducer, while an optimum blur is found for small signals.     

Amorphous selenium flat panel detectors for digital mammography: validation of a NPWE model observer with CDMAM observer performance experiments

Model observers have been developed which incorporate a specific imaging task, system performance, and human observer characteristics and can potentially overcome some of the limitations in using detective quantum efficiency for optimization and comparison of detectors. In this paper, a modified nonprewhitening matched filter (NPWE) model observer was developed and validated to predict object detectability for an amorphous selenium (a-Se) direct flat-panel imager (FPI) where aliasing is severe. A preclinical a-Se digital mammography FPI with 85 microm pixel size was used in this investigation. Its physical imaging properties including modulation transfer function (MTF), noise power spectrum, and DQE were fully characterized. An observer performance study was conducted by imaging the CDMAM 3.4 contrast-detail phantom designed specifically for digital mammography and presenting these images to a panel of seven observers. X-ray attenuation and scatter due to the phantom were determined experimentally for use in development of the model observer. The observer study results were analyzed via threshold averaging and signal detection theory (SDT) based techniques to produce contrast-detail curves where threshold contrast is plotted as a function of disk diameter. Validity of the model was established using SDT analysis of the experimental data. The effect of aliasing on the detectability of small diameter disks was determined using the NPWE model observer. The signal spectrum was calculated using the presampling MTF of the detector with and without including the aliased terms. Our results indicate that the NPWE model based on Fourier domain parameters provides reasonable prediction of object detectability for the signal-known-exactly task in uniform image noise for a-Se direct FPI.     

La théorie statistique de la décision: application à l''étude du rapport signal sur bruit en tomographie d''émission monophotoniqueStatistical decision theory: application to the study of signal to noise ratio in single photon emission tomography

The statistical decision theory (SDT) is applied in all direct imaging modalities. The detection stage can be quantified exactly from the ideal observer concept. The ideal observer depend on the noise power spectrum (NPS). An expression of NPS in single photon emission tomography (SPECT) of the uncertainty over the number of counts is derived in order to estimate the noise and the signal to noise ratio (SNR). The reconstruction algorithm is the method of back-projection of filtered projection. It is shown that derived expressions of noise and SNR compare well with expressions derived from classical approaches. So, we can affirm that the SDT can be considered in tomography (indirect imaging).     

Rapid calculation of detectability in Bayesian single photon emission computed tomography

We consider the calculation of lesion detectability using a mathematical model observer, the channelized Hotelling observer (CHO), in a signal-known-exactly/background-known-exactly detection task for single photon emission computed tomography (SPECT). We focus on SPECT images reconstructed with Bayesian maximum a posteriori methods. While model observers are designed to replace time-consuming studies using human observers, the calculation of CHO detectability is usually accomplished using a large number of sample images, which is still time consuming. We develop theoretical expressions for a measure of detectability, the signal-to-noise-ratio (SNR) of a CHO observer, that can be very rapidly evaluated. Key to our expressions are approximations to the reconstructed image covariance. In these approximations, we use methods developed in the PET literature, but modify them to reflect the different nature of attenuation and distance-dependent blur in SPECT. We validate our expressions with Monte Carlo methods. We show that reasonably accurate estimates of the SNR can be obtained at a computational expense equivalent to approximately two projection operations, and that evaluating SNR for subsequent lesion locations requires negligible additional computation.     

Evaluating the impact of X-ray spectral shape on image quality in flat-panel CT breast imaging

In recent years, there has been an increasing interest in exploring the feasibility of dedicated computed tomography (CT) breast imaging using a flat-panel digital detector in a truncated cone-beam imaging geometry. Preliminary results are promising and it appears as if three-dimensional tomographic imaging of the breast has great potential for reducing the masking effect of superimposed parenchymal structure typically observed with conventional mammography. In this study, a mathematical framework used for determining optimal design and acquisition parameters for such a CT breast imaging system is described. The ideal observer signal-to-noise ratio (SNR) is used as a figure of merit, under the assumptions that the imaging system is linear and shift invariant. Computation of the ideal observer SNR used a parallel-cascade model to predict signal and noise propagation through the detector, as well as a realistic model of the lesion detection task in breast imaging. For all evaluations, the total mean glandular dose for a CT breast imaging study was constrained to be approximately equivalent to that of a two-view conventional mammography study. The framework presented was used to explore the effect of x-ray spectral shape across an extensive range of kVp settings, filter material types, and filter thicknesses. The results give an indication of how spectral shape can affect image quality in flat-panel CT breast imaging.     

The evaluation of morphological characteristics of human spermatozoa according to stricter criteria

The evaluation of the morphology of human spermatozoa varies widely between and sometimes even within laboratories. The purpose of this study was to determine whether the method that has been developed in our laboratory and which resulted in the use of stricter criteria for the evaluation of sperm morphology is a practical, reliable and repeatable method and to establish the within and between observer variations. The criteria used for a 'normal' spermatozoon are based on the appearance of spermatozoa found in the mucus of the upper endocervical canal. The results of the morphological evaluations of 26 samples by four observers were statistically analysed by various methods. The method of Barnett showed a high degree of relative accuracy between observers with error variances of between 2.89 and 19.67 as well as high Spearman rank correlation coefficients of between 0.8675 and 0.6537 (P less than 0.0003). The Spearman correlation coefficient for 15 duplicate evaluations by one observer was 0.9650 (P less than 0.0001) while the coefficients of variation for repeated evaluations of single samples were also within acceptable limits. Based on these results, the method described in this article allows comparable and reliable results between and within observers to be obtained. From this and other studies it can be concluded that the method also has a good prognostic value for the prediction of expected IVF fertilization, the hamster test and hemizona assay.     

Observer variation in the histological grading of rectal carcinoma.

The variation between two observers in grading 100 biopsies and the corresponding main specimens of rectal carcinomas has been examined. Using kappa statistics, which take account of chance agreement, we found a highly significant level of agreement. As expected, higher levels were obtained for intraobserver agreement. However, disagreements between observers were in many instances "haphazard" and there were differences in bias between them. Fifty paired biopsies and main tumours were graded by five observers and the results analysed for bias and by kappa statistics for overall and conditional agreement. These methods revealed significant overall agreement but the levels for some observer pairs did not differ significantly from chance. Examination for observer bias indicated differing standards of grading, and haphazard disagreements reached high levels for some observer pairs. The intraobserver agreement between the grade of the biopsy and the corresponding main tumour varied from 56-69% but only 52% of the poorly differentiated tumours were diagnosed as such in the preoperative biopsy by the "specialist" observer. The poor predictive value was not improved by taking multiple biopsies. We conclude that the grade of a rectal carcinoma cannot be accurately assessed on a preoperative biopsy and that this has serious implications for the management of low rectal cancers. Furthermore the wide discrepancies in diagnostic standards between some pathologists mean that studies on the treatment and prognosis of rectal cancer which utilise histological grade for comparison purposes must be viewed with considerable skepticism.     

Image quality control for digital mammographic systems: Initial experience and outlook

This report presents (1) a broad topical review and a tutorial of the possibilities for image quality control (IQC) with digital systems, and (2) results and initial experience for IQC with two commercial digital imaging systems, but with limited discussion on any particular method. Digital imaging systems used for mammographically guided digital stereotactic breast biopsy were evaluated extensively at the University of Arizona. Measurements were made of linearity, sensitivity, signal-to-noise ratio, and square-wave modulation. Images of phantoms such as the American College of Radiology Accreditation Phantom and the contrast detail mammography Phantom were evaluated as well as images of the x-ray source’s focal spot. The evaluation also included the cathode ray tubes for the imaging systems. The data collected show that digital imaging systems have an important advantage over film-screen systems because they provide a digital signal as output that can be used for quantitative analysis. As a result, IQC can become a much more quantitative discipline than presently practiced, providing more information on the imaging systems under evaluation, and providing better control over their properties during actual operation.     

Optimization of a tomosynthesis system for the detection of lung nodules

Mathematical observers that track human performance can be used to reduce the number of human observer studies needed to optimize imaging systems. The performance of human observers for the detection of a 3.6 mm lung nodule in anatomical backgrounds was measured as a function of varying tomosynthetic angle and compared with mathematical observers. The human observer results showed a dramatic increase in the percent of correct responses, from 80% in the projection images to 96% in the projection images with a tomosynthetic angle of just 3 degrees. This result suggests the potential usefulness of the scanned beam digital x-ray system for this application. Given the small number of images (40) used per tomosynthetic angle and the highly nonstationary statistical nature of the backgrounds, the nonprewhitening eye observer achieved a higher performance than the channelized Hotelling observer using a Laguerre-Gauss basis. The channelized Hotelling observer with internal noise and the eye filter matched to the projection data were shown to track human performance as the tomosynthetic angle changed. The validation of these mathematical observers extends their applicability to the optimization of tomosynthesis systems.     

Fast LROC analysis of Bayesian reconstructed emission tomographic images using model observers

Lesion detection and localization is an important task in emission computed tomography. Detection and localization performance with signal location uncertainty may be summarized by a scalar figure of merit, the area under the localization receiver operating characteristic (LROC) curve, A(LROC). We consider model observers to compute A(LROC) for two-dimensional maximum a posteriori (MAP) reconstructions. Model observers may be used to rapidly prototype studies that use human observers. We address the case background-known-exactly (BKE) and signal known except for location. Our A(LROC) calculation makes use of theoretical expressions for the mean and covariance of the reconstruction and, unlike conventional methods that also use model observers, does not require computation of a large number of sample reconstructions. We validate the results of the procedure by comparison to A(LROC) obtained using a gold-standard Monte Carlo method employing a large set of reconstructed noise samples. Under reasonable simulation conditions, our theoretical calculation is about one to two orders of magnitude faster than the conventional Monte Carlo method.     

Assessment of analysis-of-variance-based methods to quantify the random variations of observers in medical imaging measurements: guidelines to the investigator

The random variations of observers in medical imaging measurements negatively affect the outcome of cancer treatment, and should be taken into account during treatment by the application of safety margins that are derived from estimates of the random variations. Analysis-of-variance- (ANOVA-) based methods are the most preferable techniques to assess the true individual random variations of observers, but the number of observers and the number of cases must be taken into account to achieve meaningful results. Our aim in this study is twofold. First, to evaluate three representative ANOVA-based methods for typical numbers of observers and typical numbers of cases. Second, to establish guidelines to the investigator to determine which method, how many observers, and which number of cases are required to obtain the a priori chosen performance. The ANOVA-based methods evaluated in this study are an established technique (pairwise differences method: PWD), a new approach providing additional statistics (residuals method: RES), and a generic technique that uses restricted maximum likelihood (REML) estimation. Monte Carlo simulations were performed to assess the performance of the ANOVA-based methods, which is expressed by their accuracy (closeness of the estimates to the truth), their precision (standard error of the estimates), and the reliability of their statistical test for the significance of a difference in the random variation of an observer between two groups of cases. The highest accuracy is achieved using REML estimation, but for datasets of at least 50 cases or arrangements with 6 or more observers, the differences between the methods are negligible, with deviations from the truth well below +/-3%. For datasets up to 100 cases, it is most beneficial to increase the number of cases to improve the precision of the estimated random variations, whereas for datasets over 100 cases, an improvement in precision is most efficiently achieved by increasing the number of observers. For datasets of at least 50 cases, the standard error ranges between 30% or less with 3 observers down to 10% or less with 8 observers, and the differences in precision between the methods are negligible. The F test (PWD) is very anticonservative and should not be used, while the t test (RES) is reliable for datasets of at least 2 x 50 cases evaluated by 4 or more observers. The likelihood-ratio-test (REML estimation) consistently indicates the significance of a difference in the random variation of an observer between two groups of cases, regardless of the number of cases, and regardless of the number of observers. If a statistical package to perform REML estimation is available, and the investigator feels confident using it, this is the preferred method for studies that involve less than 50 cases evaluated by less than 6 observers. Otherwise, the RES method is an excellent alternative, because of its straightforward implementation, its completeness with respect to the provided statistics, and its overall sufficient accuracy, precision, and reliability of the provided statistical test. If neither the RES method nor REML estimation can provide sufficient performance, either more observers or more cases must be included.     

Quantification of diverse subcellular immunohistochemical markers with clinicobiological relevancies: validation of a new computer-assisted image analysis procedure

Tissue microarray technology and immunohistochemical techniques have become a routine and indispensable tool for current anatomical pathology diagnosis. However, manual quantification by eye is relatively slow and subjective, and the use of digital image analysis software to extract information of immunostained specimens is an area of ongoing research, especially when the immunohistochemical signals have different localization in the cells (nuclear, membrane, cytoplasm). To minimize critical aspects of manual quantitative data acquisition, we generated semi-automated image-processing steps for the quantification of individual stained cells with immunohistochemical staining of different subcellular location. The precision of these macros was evaluated in 196 digital colour images of different Hodgkin lymphoma biopsies stained for different nuclear (Ki67, p53), cytoplasmic (TIA-1, CD68) and membrane markers (CD4, CD8, CD56, HLA-Dr). Semi-automated counts were compared to those obtained manually by three separate observers. Paired t -tests demonstrated significant differences between intra- and inter-observer measurements, with more substantial variability when the cellular density of the digital images was > 100 positive cells/image. Overall, variability was more pronounced for intra-observer than for inter-observer comparisons, especially for cytoplasmic and membrane staining patterns ( P  < 0.0001 and P  = 0.050). The comparison between the semi-automated and manual microscopic measurement methods indicates significantly lower variability in the results yielded by the former method. Our semi-automated computerized method eliminates the major causes of observer variability and may be considered a valid alternative to manual microscopic quantification for diagnostic, prognostic and therapeutic purposes.     

Automatic quantitative low contrast analysis of digital chest phantom radiographs

Up to the present, the majority of low contrast object evaluations performed on phantom images have been accomplished in a subjective fashion. This is mainly due to the time and effort required to manually measure each radiograph with a densitometer to obtain quantitative results. However, with the development of digital radiographic systems, it has become feasible to automate the detection and computation processes. In this work, a method that can automatically detect and compute the subject-to-noise ratio of the low contrast disks inside a geometric chest phantom is examined. This algorithm has the ability to locate the low contrast objects to an accuracy of less than one pixel, and provides results that are consistent with the understanding of subject-to-noise ratio. This algorithm should simplify the task of quantitative evaluation of contrast detail phantoms.     

Seashore Rhythm test: comparison of Signal Detection Theory and standard scoring procedures.

The effects of age and instructional set on Seashore Rhythm test performance were examined with standard scoring and Signal Detection Theory (SDT) procedures. Neurologically intact young (17-28 years) and old (65-92 years) subjects were given standard test instructions with sentences added which stated that specific proportions of the test stimuli would be identical or different. Age differences in d' scores were highly correlated with standard Rhythm scores. While standard scores were unaffected by instructional set, all three SDT measures of response criteria detected effects of instructional set, or an age by instructional set interaction. SDT scoring of the Rhythm test yields a measure of auditory sensitivity comparable to standard methods and detects response bias on the lest that standard procedures cannot.     

Evaluation of internal noise methods for Hotelling observer models

The inclusion of internal noise in model observers is a common method to allow for quantitative comparisons between human and model observer performance in visual detection tasks. In this article, we studied two different strategies for inserting internal noise into Hotelling model observers. In the first strategy, internal noise was added to the output of individual channels: (a) Independent nonuniform channel noise, (b) independent uniform channel noise. In the second strategy, internal noise was added to the decision variable arising from the combination of channel responses. The standard deviation of the zero mean internal noise was either constant or proportional to: (a) the decision variable's standard deviation due to the external noise, (b) the decision variable's variance caused by the external noise, (c) the decision variable magnitude on a trial to trial basis. We tested three model observers: square window Hotelling observer (HO), channelized Hotelling observer (CHO), and Laguerre-Gauss Hotelling observer (LGHO) using a four alternative forced choice (4AFC) signal known exactly but variable task with a simulated signal embedded in real x-ray coronary angiogram backgrounds. The results showed that the internal noise method that led to the best prediction of human performance differed across the studied model observers. The CHO model best predicted human observer performance with the channel internal noise. The HO and LGHO best predicted human observer performance with the decision variable internal noise. The present results might guide researchers with the choice of methods to include internal noise into Hotelling model observers when evaluating and optimizing medical image quality.     

A method for energy window optimization for quantitative tasks that includes the effects of model-mismatch on bias: application to Y-90 bremsstrahlung SPECT imaging

Quantitative Yttrium-90 ((90)Y) bremsstrahlung single photon emission computed tomography (SPECT) imaging has shown great potential to provide reliable estimates of (90)Y activity distribution for targeted radionuclide therapy dosimetry applications. One factor that potentially affects the reliability of the activity estimates is the choice of the acquisition energy window. In contrast to imaging conventional gamma photon emitters where the acquisition energy windows are usually placed around photopeaks, there has been great variation in the choice of the acquisition energy window for (90)Y imaging due to the continuous and broad energy distribution of the bremsstrahlung photons. In quantitative imaging of conventional gamma photon emitters, previous methods for optimizing the acquisition energy window assumed unbiased estimators and used the variance in the estimates as a figure of merit (FOM). However, for situations, such as (90)Y imaging, where there are errors in the modeling of the image formation process used in the reconstruction there will be bias in the activity estimates. In (90)Y bremsstrahlung imaging this will be especially important due to the high levels of scatter, multiple scatter, and collimator septal penetration and scatter. Thus variance will not be a complete measure of reliability of the estimates and thus is not a complete FOM. To address this, we first aimed to develop a new method to optimize the energy window that accounts for both the bias due to model-mismatch and the variance of the activity estimates. We applied this method to optimize the acquisition energy window for quantitative (90)Y bremsstrahlung SPECT imaging in microsphere brachytherapy. Since absorbed dose is defined as the absorbed energy from the radiation per unit mass of tissues in this new method we proposed a mass-weighted root mean squared error of the volume of interest (VOI) activity estimates as the FOM. To calculate this FOM, two analytical expressions were derived for calculating the bias due to model-mismatch and the variance of the VOI activity estimates, respectively. To obtain the optimal acquisition energy window for general situations of interest in clinical (90)Y microsphere imaging, we generated phantoms with multiple tumors of various sizes and various tumor-to-normal activity concentration ratios using a digital phantom that realistically simulates human anatomy, simulated (90)Y microsphere imaging with a clinical SPECT system and typical imaging parameters using a previously validated Monte Carlo simulation code, and used a previously proposed method for modeling the image degrading effects in quantitative SPECT reconstruction. The obtained optimal acquisition energy window was 100-160?keV. The values of the proposed FOM were much larger than the FOM taking into account only the variance of the activity estimates, thus demonstrating in our experiment that the bias of the activity estimates due to model-mismatch was a more important factor than the variance in terms of limiting the reliability of activity estimates.     

SNR Weighting for Shear Wave Speed Reconstruction in Tomoelastography

In tomoelastography, to achieve a final wave speed map by combining reconstructions obtained from all spatial directions and excitation frequencies, the use of weights is inevitable. Here, a new weighting scheme, which maximizes the signal‐to‐noise ratio (SNR) of the final wave speed map, has been proposed. To maximize the SNR of the final wave speed map, the use of squares of estimated SNR values of reconstructed individual maps has been proposed. Therefore, derivations of the SNR of the reconstructed wave speed maps have become necessary. Considering the noise on the complex MRI signal, the SNR of the reconstructed wave speed map was formulated by an analytical approach assuming a high SNR, and the results were verified using Monte Carlo simulations (MCSs). It has been assumed that the noise remains approximately Gaussian when the image SNR is high enough, despite the nonlinear operations in tomoelastography inversion. Hence, the SNR threshold was determined by comparing the SNR computed by MCSs and analytical approximations. The weighting scheme was evaluated for accuracy, spatial resolution and SNR performances on simulated phantoms. MR elastography (MRE) experiments on two different phantoms were conducted. Wave speed maps were generated for simulated 3D human abdomen MRE data and experimental human abdomen MRE data. The simulation results demonstrated that the SNR‐weighted inversion improved the SNR performance of the wave speed map by a factor of two compared to the performance of the original (i.e., amplitude‐weighted) reconstruction. In the case of a low SNR, no bias occurred in the wave speed map when SNR weighting was used, whereas 10% bias occurred when the original weighting (i.e., amplitude weighting) was used. Thus, while not altering the accuracy or spatial resolution of the wave speed map with the proposed weighting method, the SNR of the wave speed map has been significantly improved.     

Comparison of region-of-interest analysis and human observers in the diagnosis of Parkinson's disease using [99mTc]TRODAT-1 and SPECT

This study determined the relative accuracy of diagnosis of Parkinson's disease (PD) using SPECT imaging data, comparing a semi-quantitative region-of-interest (ROI) approach and human observers. A set of patients with PD and normal healthy control subjects were studied using the dopamine transporter tracer [(99m)Tc]TRODAT-1 and SPECT. The sample comprised 81 patients (mean age +/- SD, 63.4 +/- 10.4 years; age range, 39.0-84.2 years) and 94 healthy controls (mean age +/- SD, 61.8 +/- 11.0 years; age range, 40.9-83.3 years). A standardized template containing six ROIs was transposed onto subregions of the brain, and the ratio of striatal to background ROI values was used as a semi-quantitative outcome measure. All images were used in a human observer study, with four experienced investigators. The data from the observer and ROI studies were analysed using a receiver operating characteristic (ROC) analysis, where the area under the ROC curve (AUC) indicated the diagnostic accuracy. ROI analysis and human observers gave similar diagnostic performance (mean observer AUC = 0.89, best ROI AUC = 0.90). This suggested that the human observers are visually acquiring similar information from the images that are contained in the semi-quantitative striatal uptake.