Microcalcification detectability in breast CT images using CNN observers

BACKGROUND

Breast computed tomography (CT) is an emerging breast imaging modality, and ongoing developments aim to improve breast CT's ability to detect microcalcifications. To understand the effects of different parameters on microcalcification detectability, a virtual clinical trial study was conducted using hybrid images and convolutional neural network (CNN)-based model observers. Mathematically generated microcalcifications were embedded into breast CT data sets acquired at our institution, and parameters related to calcification size, calcification contrast, cluster diameter, cluster density, and image display method (i.e., single slices, slice averaging, and maximum-intensity projections) were evaluated for their influence on microcalcification detectability.

PURPOSE

To investigate the individual effects and the interplay of parameters affecting microcalcification detectability in breast CT.

METHODS

Spherical microcalcifications of varying diameters (0.04, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40 mm) and native intensities were computer simulated to portray the partial volume effects of the imaging system. Calcifications were mathematically embedded into 109 patient breast CT volume data sets as individual calcifications or as clusters of calcifications. Six numbers of calcifications (1, 3, 5, 7, 10, 15) distributed within six cluster diameters (1, 3, 5, 6, 8, 10 mm) were simulated to study the effect of cluster density. To study the role of image display method, 2D regions of interest (ROIs) and 3D volumes of interest (VOIs) were generated using single slice extraction, slice averaging, and maximum-intensity projection (MIP). 2D and 3D CNNs were trained on the ROIs and VOIs, and receiver operating characteristic (ROC) curve analysis was used to evaluate detection performance. The area under the ROC curve (AUC) was used as the primary performance metric.

RESULTS

Detection performance decreased with increasing section thickness, and peak detection performance occurred using the native section thickness (0.2 mm) and MIP display. The MIP display method, despite using a single slice, yielded comparable performance to the native section thickness, which employed 50 slices. Reduction in slices did not sacrifice detection accuracy and provided significant computational advantages over multi-slice image volumes. Larger cluster diameters resulted in reduced overall detectability, while smaller cluster diameters led to increased detectability. Additionally, we observed that the presence of more calcifications within a cluster improved the overall detectability, while fewer calcifications decreased it.

CONCLUSIONS

As breast CT is still a relatively new breast imaging modality, there is an ongoing need to identify optimal imaging protocols. This work demonstrated the utility of MIP presentation for displaying image volumes containing microcalcification clusters. It is likely that human observers may also benefit from viewing MIPs compared to individual slices. The results of this investigation begin to elucidate how model observers interact with microcalcification clusters in a 3D volume, and will be useful for future studies investigating a broader set of parameters related to breast CT.

Pre-whitened matched filter and convolutional neural network based model observer performance for mass lesion detection in non-contrast breast CT

BACKGROUND

Mathematical model observers have been shown to reasonably predict human observer performance and are useful when human observer studies are infeasible. Recently, convolutional neural networks (CNNs) have also been used as substitutes for human observers, and studies have shown their utility as an optimal observer. In this study, a CNN model observer is compared to the pre-whitened matched filter (PWMF) model observer in detecting simulated mass lesions inserted into 253 acquired breast computed tomography (bCT) images from patients imaged at our institution.

PURPOSE

To compare CNN and PWMF model observers for detecting signal-known-exactly (SKE) location-known-exactly (LKE) simulated lesions in bCT images with real anatomical backgrounds, and to use these model observers collectively to optimize parameters and understand trends in performance with breast CT.

METHODS

Spherical lesions with different diameters (1, 3, 5, 9 mm) were mathematically inserted into reconstructed patient bCT image data sets to mimic 3D mass lesions in the breast. 2D images were generated by extracting the center slice along the axial dimension or by slice averaging across adjacent slices to model thicker sections (0.4, 1.2, 2.0, 6.0, 12.4, 20.4 mm). The role of breast density was retrospectively studied using the range of breast densities intrinsic to the patient bCT data sets. In addition, mass lesions were mathematically inserted into Gaussian images matched to the mean and noise power spectrum of the bCT images to better understand the performance of the CNN in the context of a known ideal observer (the PWMF). The simulated Gaussian and bCT images were divided into training and testing data sets. Each training data set consisted of 91 600 images, and each testing data set consisted of 96 000 images. A CNN and PWMF was trained on the Gaussian training images, and a different CNN and PWMF was trained on the bCT training images. The trained model observers were tested, and receiver operating characteristic (ROC) curve analysis was used to evaluate detection performance. The area under the ROC curve (AUC) was the primary performance metric used to compare the model observers.

RESULTS

In the Gaussian background, the CNN performed essentially identically to the PWMF across lesion sizes and section thicknesses. In the bCT background, the CNN outperformed the PWMF across lesion size, breast density, and most section thicknesses. These findings suggest that there are higher-order features in bCT images that are harnessed by the CNN observer but are inaccessible to the PWMF.

CONCLUSIONS

The CNN performed equivalently to the ideal observer in Gaussian textures. In bCT background, the CNN captures more diagnostic information than the PWMF and may be a more pertinent observer when conducting optimal performance studies in breast CT images.

Model observer performance in contrast-enhanced lesions in breast CT: The influence of contrast concentration on detectability

BACKGROUND

The use of iodine-based contrast agent for better delineation of tumors in breast CT (bCT) has been shown to be compelling, similar to the tumor enhancement in contrast-enhanced breast MRI. Contrast-enhanced bCT (CE-bCT) is a relatively new tool, and a structured evaluation of different imaging parameters at play has yet to be conducted. In this investigation, data sets of acquired bCT images from 253 patients imaged at our institution were used in concert with simulated mathematically inserted spherical contrast-enhanced lesions to study the role of contrast enhancement on detectability.

PURPOSE

To quantitatively evaluate the improvement in lesion detectability due to contrast enhancement across lesion diameter, section thickness, view plane, and breast density using a pre-whitened matched filter (PWMF) model observer.

METHODS

The relationship between iodine concentration and Hounsfield units (HU) was measured using spectral modeling. The lesion enhancement from clinical CE-bCT images in 22 patients was evaluated, and the average contrast enhancement (ΔHU) was determined. Mathematically generated spherical mass lesions of varying diameters (1, 3, 5, 9, 11, 15 mm) and contrast enhancement levels (0, 0.25, 0.50, 0.75, 1) were inserted at random locations in 253 actual patient bCT datasets. Images with varying thicknesses (0.4-19.8 mm) were generated by slice averaging, and the role of view plane (coronal and axial planes) was studied. A PWMF was used to generate receiver operating characteristic (ROC) curves across parameters of lesion diameter, contrast enhancement, section thickness, view plane, and breast density. The area under the ROC curve (AUC) was used as the primary performance metric, generated from over 90,000 simulated lesions.

RESULTS

An average 20% improvement (ΔAUC = 0.1) in lesion detectability due to contrast enhancement was observed across lesion diameter, section thickness, breast density, and view plane. A larger improvement was observed when stratifying patients based on breast density. For patients with VGF ≤ 40%, detection performance improved up to 20% (until AUC →1), and for patients with denser breasts (VGF > 40%), detection performance improved more drastically, ranging from 20% to 80% for 1- and 5-mm lesions. For the 1 mm lesion, detection performance raised slightly at the 1.2 mm section thickness before falling off as thickness increased. For larger lesions, detection performance was generally unaffected as section thickness increased up until it reached 5.8 mm, where performance began to decline. Detection performance was higher in the axial plane compared to the coronal plane for smaller lesions and thicker sections.

CONCLUSIONS

For emerging diagnostic tools like CE-bCT, it is important to optimize imaging protocols for lesion detection. In this study, we found that intravenous contrast can be used to detect small lesions in dense breasts. Optimal section thickness for detectability has dependencies on breast density and lesion size, therefore, display thickness should be adjusted in real-time using display software. These findings may be useful for the development of CE-bCT as well as other x-ray-based breast imaging modalities.

Local texture statistics augment the power spectrum in modeling radiographic judgments of breast density

Purpose

Anatomical "noise" is an important limitation of full-field digital mammography. Understanding its impact on clinical judgments is made difficult by the complexity of breast parenchyma, which results in image texture not fully captured by the power spectrum. While the number of possible parameters for characterizing anatomical noise is quite large, a specific set of local texture statistics has been shown to be visually salient, and human sensitivity to these statistics corresponds to their informativeness in natural scenes.

Approach

We evaluate these local texture statistics in addition to standard power-spectral measures to determine whether they have additional explanatory value for radiologists' breast density judgments. We analyzed an image database consisting of 111 disease-free mammographic screening exams (4 views each) acquired at the University of Pittsburgh Medical Center. Each exam had a breast density score assigned by the examining radiologist. Power-spectral descriptors and local image statistics were extracted from images of breast parenchyma. Model-selection criteria and accuracy were used to assess the explanatory and predictive value of local image statistics for breast density judgments.

Results

The model selection criteria show that adding local texture statistics to descriptors of the power spectra produce better explanatory and predictive models of radiologists' judgments of breast density. Thus, local texture statistics capture, in some form, non-Gaussian aspects of texture that radiologists are using.

Conclusions

Since these local texture statistics are expected to be impacted by imaging factors like modality, dose, and image processing, they suggest avenues for understanding and optimizing observer performance.

DEMIST: A deep-learning-based task-specific denoising approach for myocardial perfusion SPECT

There is an important need for methods to process myocardial perfusion imaging (MPI) SPECT images acquired at lower radiation dose and/or acquisition time such that the processed images improve observer performance on the clinical task of detecting perfusion defects. To address this need, we build upon concepts from model-observer theory and our understanding of the human visual system to propose a Detection task-specific deep-learning-based approach for denoising MPI SPECT images (DEMIST). The approach, while performing denoising, is designed to preserve features that influence observer performance on detection tasks. We objectively evaluated DEMIST on the task of detecting perfusion defects using a retrospective study with anonymized clinical data in patients who underwent MPI studies across two scanners (N = 338). The evaluation was performed at low-dose levels of 6.25%, 12.5% and 25% and using an anthropomorphic channelized Hotelling observer. Performance was quantified using area under the receiver operating characteristics curve (AUC). Images denoised with DEMIST yielded significantly higher AUC compared to corresponding low-dose images and images denoised with a commonly used task-agnostic DL-based denoising method. Similar results were observed with stratified analysis based on patient sex and defect type. Additionally, DEMIST improved visual fidelity of the low-dose images as quantified using root mean squared error and structural similarity index metric. A mathematical analysis revealed that DEMIST preserved features that assist in detection tasks while improving the noise properties, resulting in improved observer performance. The results provide strong evidence for further clinical evaluation of DEMIST to denoise low-count images in MPI SPECT.

Optimizing the Pairs of Radiologists That Double Read Screening Mammograms

Background Despite variation in performance characteristics among radiologists, the pairing of radiologists for the double reading of screening mammograms is performed randomly. It is unknown how to optimize pairing to improve screening performance. Purpose To investigate whether radiologist performance characteristics can be used to determine the optimal set of pairs of radiologists to double read screening mammograms for improved accuracy. Materials and Methods This retrospective study was performed with reading outcomes from breast cancer screening programs in Sweden (2008-2015), England (2012-2014), and Norway (2004-2018). Cancer detection rates (CDRs) and abnormal interpretation rates (AIRs) were calculated, with AIR defined as either reader flagging an examination as abnormal. Individual readers were divided into performance categories based on their high and low CDR and AIR. The performance of individuals determined the classification of pairs. Random pair performance, for which any type of pair was equally represented, was compared with the performance of specific pairing strategies, which consisted of pairs of readers who were either opposite or similar in AIR and/or CDR. Results Based on a minimum number of examinations per reader and per pair, the final study sample consisted of 3 592 414 examinations (Sweden, n = 965 263; England, n = 837 048; Norway, n = 1 790 103). The overall AIRs and CDRs for all specific pairing strategies (Sweden AIR range, 45.5-56.9 per 1000 examinations and CDR range, 3.1-3.6 per 1000; England AIR range, 68.2-70.5 per 1000 and CDR range, 8.9-9.4 per 1000; Norway AIR range, 81.6-88.1 per 1000 and CDR range, 6.1-6.8 per 1000) were not significantly different from the random pairing strategy (Sweden AIR, 54.1 per 1000 examinations and CDR, 3.3 per 1000; England AIR, 69.3 per 1000 and CDR, 9.1 per 1000; Norway AIR, 84.1 per 1000 and CDR, 6.3 per 1000). Conclusion Pairing a set of readers based on different pairing strategies did not show a significant difference in screening performance when compared with random pairing. © RSNA, 2023.

Poster Session: A direct measure of adaptation and visual salience

One hypothesized function of adaptation is to increase the salience of novel targets by discounting the properties of the ambient environment. Previous studies have suggested this by finding faster search times for novel targets when searching on backgrounds observers are currently adapted to. However, this provides only an indirect measure of salience. Here, we developed a more direct measure of the impact of adaptation on feature salience. Backgrounds were oriented 1/f noise images with power confined within 15 deg of horizontal or vertical. Targets were 5 c/deg Gabor patches centered on the 8 deg backgrounds. Observers simultaneously adapted to the horizontal or vertical backgrounds shown on the left or right of fixation. A 250ms test probe then showed the Gabor patch on the same background (horizontal or vertical) on both sides. The target orientation was adjusted on one side until it appeared as conspicuous as a fixed target on the other side. Settings were made for fixed targets ranging from 10 to 45 deg from the backgrounds. For most conditions/observers, the salience matches required a smaller orientation offset on the same- vs. different-adapt background. These results support a functional role of adaptation in highlighting novelty by potentially "unmasking" the target from its background, and emphasize the importance of considering adaptation aftereffects not only for isolated targets but within the stimulus contexts they are embedded in.

A 2D Synthesized Image Improves the 3D Search for Foveated Visual Systems

Current medical imaging increasingly relies on 3D volumetric data making it difficult for radiologists to thoroughly search all regions of the volume. In some applications (e.g., Digital Breast Tomosynthesis), the volumetric data is typically paired with a synthesized 2D image (2D-S) generated from the corresponding 3D volume. We investigate how this image pairing affects the search for spatially large and small signals. Observers searched for these signals in 3D volumes, 2D-S images, and while viewing both. We hypothesize that lower spatial acuity in the observers' visual periphery hinders the search for the small signals in the 3D images. However, the inclusion of the 2D-S guides eye movements to suspicious locations, improving the observer's ability to find the signals in 3D. Behavioral results show that the 2D-S, used as an adjunct to the volumetric data, improves the localization and detection of the small (but not large) signal compared to 3D alone. There is a concomitant reduction in search errors as well. To understand this process at a computational level, we implement a Foveated Search Model (FSM) that executes human eye movements and then processes points in the image with varying spatial detail based on their eccentricity from fixations. The FSM predicts human performance for both signals and captures the reduction in search errors when the 2D-S supplements the 3D search. Our experimental and modeling results delineate the utility of 2D-S in 3D search-reduce the detrimental impact of low-resolution peripheral processing by guiding attention to regions of interest, effectively reducing errors.

Monitoring Muscle Perfusion in Rodents During Short-Term Ischemia Using Power Doppler Ultrasound

OBJECTIVE

The aim of this work was to evaluate the reliability of power Doppler ultrasound (PD-US) measurements made without contrast enhancement to monitor temporal changes in peripheral blood perfusion.

METHODS

On the basis of pre-clinical rodent studies, we found that combinations of spatial registration and clutter filtering techniques applied to PD-US signals reproducibly tracked blood perfusion in skeletal muscle. Perfusion is monitored while modulating hindlimb blood flow. First, in invasive studies, PD-US measurements in deep muscle with laser speckle contrast imaging (LSCI) of superficial tissues made before, during and after short-term arterial clamping were compared. Then, in non-invasive studies, a pressure cuff was employed to generate longer-duration hindlimb ischemia. Here, B-mode imaging was also applied to measure flow-mediated dilation of the femoral artery while, simultaneously, PD-US was used to monitor downstream muscle perfusion to quantify reactive hyperemia. Measurements in adult male and female mice and rats, some with exercise conditioning, were included to explore biological variables.

RESULTS

PD-US methods are validated through comparisons with LSCI measurements. As expected, no significant differences were found between sexes or fitness levels in flow-mediated dilation or reactive hyperemia estimates, although post-ischemic perfusion was enhanced with exercise conditioning, suggesting there could be differences between the hyperemic responses of conduit and resistive vessels.

CONCLUSION

Overall, we found non-contrast PD-US imaging can reliably monitor relative spatiotemporal changes in muscle perfusion. This study supports the development of PD-US methods for monitoring perfusion changes in patients at risk for peripheral artery disease.

Discrimination tasks in simulated low-dose CT noise

BACKGROUND

This study reports the results of a set of discrimination experiments using simulated images that represent the appearance of subtle lesions in low-dose computed tomography (CT) of the lungs. Noise in these images has a characteristic ramp-spectrum before apodization by noise control filters. We consider three specific diagnostic features that determine whether a lesion is considered malignant or benign, two system-resolution levels, and four apodization levels for a total of 24 experimental conditions.

PURPOSE

The goal of the investigation is to better understand how well human observers perform subtle discrimination tasks like these, and the mechanisms of that performance. We use a forced-choice psychophysical paradigm to estimate observer efficiency and classification images. These measures quantify how effectively subjects can read the images, and how they use images to perform discrimination tasks across the different imaging conditions.

MATERIALS AND METHODS

The simulated CT images used as stimuli in the psychophysical experiments are generated from high-resolution objects passed through a modulation transfer function (MTF) before down-sampling to the image-pixel grid. Acquisition noise is then added with a ramp noise-power spectrum (NPS), with subsequent smoothing through apodization filters. The features considered are lesion size, indistinct lesion boundary, and a nonuniform lesion interior. System resolution is implemented by an MTF with resolution (10% max.) of 0.47 or 0.58 cyc/mm. Apodization is implemented by a Shepp-Logan filter (Sinc profile) with various cutoffs. Six medically naïve subjects participated in the psychophysical studies, entailing training and testing components for each condition. Training consisted of staircase procedures to find the 80% correct threshold for each subject, and testing involved 2000 psychophysical trials at the threshold value for each subject. Human-observer performance is compared to the Ideal Observer to generate estimates of task efficiency. The significance of imaging factors is assessed using ANOVA. Classification images are used to estimate the linear template weights used by subjects to perform these tasks. Classification-image spectra are used to analyze subject weights in the spatial-frequency domain.

RESULTS

Overall, average observer efficiency is relatively low in these experiments (10%-40%) relative to detection and localization studies reported previously. We find significant effects for feature type and apodization level on observer efficiency. Somewhat surprisingly, system resolution is not a significant factor. Efficiency effects of the different features appear to be well explained by the profile of the linear templates in the classification images. Increasingly strong apodization is found to both increase the classification-image weights and to increase the mean-frequency of the classification-image spectra. A secondary analysis of "Unapodized" classification images shows that this is largely due to observers undoing (inverting) the effects of apodization filters.

CONCLUSIONS

These studies demonstrate that human observers can be relatively inefficient at feature-discrimination tasks in ramp-spectrum noise. Observers appear to be adapting to frequency suppression implemented in apodization filters, but there are residual effects that are not explained by spatial weighting patterns. The studies also suggest that the mechanisms for improving performance through the application of noise-control filters may require further investigation.

Development and task-based evaluation of a scatter-window projection and deep learning-based transmission-less attenuation compensation method for myocardial perfusion SPECT

Attenuation compensation (AC) is beneficial for visual interpretation tasks in single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI). However, traditional AC methods require the availability of a transmission scan, most often a CT scan. This approach has the disadvantages of increased radiation dose, increased scanner cost, and the possibility of inaccurate diagnosis in cases of misregistration between the SPECT and CT images. Further, many SPECT systems do not include a CT component. To address these issues, we developed a Scatter-window projection and deep Learning-based AC (SLAC) method to perform AC without a separate transmission scan. To investigate the clinical efficacy of this method, we then objectively evaluated the performance of this method on the clinical task of detecting perfusion defects on MPI in a retrospective study with anonymized clinical SPECT/CT stress MPI images. The proposed method was compared with CT-based AC (CTAC) and no-AC (NAC) methods. Our results showed that the SLAC method yielded an almost overlapping receiver operating characteristic (ROC) plot and a similar area under the ROC (AUC) to the CTAC method on this task. These results demonstrate the capability of the SLAC method for transmission-less AC in SPECT and motivate further clinical evaluation.

Development and task-based evaluation of a scatter-window projection and deep learning-based transmission-less attenuation compensation method for myocardial perfusion SPECT

Attenuation compensation (AC) is beneficial for visual interpretation tasks in single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI). However, traditional AC methods require the availability of a transmission scan, most often a CT scan. This approach has the disadvantage of increased radiation dose, increased scanner costs, and the possibility of inaccurate diagnosis in cases of misregistration between the SPECT and CT images. Further, many SPECT systems do not include a CT component. To address these issues, we developed a Scatter-window projection and deep Learning-based AC (SLAC) method to perform AC without a separate transmission scan. To investigate the clinical efficacy of this method, we then objectively evaluated the performance of this method on the clinical task of detecting perfusion defects on MPI in a retrospective study with anonymized clinical SPECT/CT stress MPI images. The proposed method was compared with CT-based AC (CTAC) and no-AC (NAC) methods. Our results showed that the SLAC method yielded an almost overlapping receiver operating characteristic (ROC) plot and a similar area under the ROC (AUC) to the CTAC method on this task. These results demonstrate the capability of the SLAC method for transmission-less AC in SPECT and motivate further clinical evaluation.

Visual adaptation to medical images: a comparison of digital mammography and tomosynthesis

Purpose

Radiologists and other image readers spend prolonged periods inspecting medical images. The visual system can rapidly adapt or adjust sensitivity to the images that an observer is currently viewing, and previous studies have demonstrated that this can lead to pronounced changes in the perception of mammogram images. We compared these adaptation effects for images from different imaging modalities to explore both general and modality-specific consequences of adaptation in medical image perception.

Approach

We measured perceptual changes induced by adaptation to images acquired by digital mammography (DM) or digital breast tomosynthesis (DBT), which have both similar and distinct textural properties. Participants (nonradiologists) adapted to images from the same patient acquired from each modality or for different patients with American College of Radiology-Breast Imaging Reporting and Data System (BI-RADS) classification of dense or fatty tissue. The participants then judged the appearance of composite images formed by blending the two adapting images (i.e., DM versus DBT or dense versus fatty in each modality).

Results

Adaptation to either modality produced similar significant shifts in the perception of dense and fatty textures, reducing the salience of the adapted component in the test images. In side-by-side judgments, a modality-specific adaptation effect was not observed. However, when the images were directly fixated during adaptation and testing, so that the textural differences between the modalities were more visible, significantly different changes in the sensitivity to the noise in the images were observed.

Conclusions

These results confirm that observers can readily adapt to the visual properties or spatial textures of medical images in ways that can bias their perception of the images, and that adaptation can also be selective for the distinctive visual features of images acquired by different modalities.

Performance of high-resolution CT for detection and discrimination tasks related to stenotic lesions - A phantom study using model observers

BACKGROUND

Accurate detection and grading of atheromatous stenotic lesions within the cardiac, renal, and intracranial vasculature is imperative for early recognition of disease and guiding treatment strategies.

PURPOSE

In this work, a stenotic lesion phantom was used to compare high resolution and normal resolution modes on the same CT scanner in terms of detection and size discrimination performance.

MATERIALS AND METHODS

The phantom is comprised of three acrylic cylinders (each 15.0 cm in diameter and 1.3 cm thick) with a matching array of holes in each module. The outer two modules contain holes that are slightly larger than the corresponding hole in the central module to simulate stenotic narrowing in vasculature. The stack of modules was submerged in an iodine solution simulating contrast-enhanced stenotic lesions with a range of lumen diameters (1.32-10.08 mm) and stenosis severity (0%, 50%, 60%, 70%, and 80%). The phantom was imaged on the Canon Aquilion Precision high-resolution CT scanner in high-resolution (HR) mode (0.25 mm × 0.50 mm detector element size) and normal-resolution (NR) mode (0.50 mm × 0.50 mm) using 120 kV and two dose levels (14 and 21 mGy SSDE) with 30 repeat scans acquired for each combination_. Filtered back-projection (FBP) and a hybrid-iterative reconstruction (AIDR) were used with the FC18 kernel, as well as a deep learning algorithm (AiCE) which is only available for HR. A non-prewhitening model observer with an eye filter was implemented to quantify performance for detection and size discrimination tasks in the axial plane.

RESULTS

Detection performance improved with increasing diameter, dose, and for AIDR in comparison to FBP for a fixed resolution mode. Performance in the HR mode was generally higher than NR for the smaller lumen diameters (1-5 mm) with decreasing differences as the diameter increased. Performance in NR mode surpassed HR mode for lumen diameters greater than ∼4 mm and ∼5 mm for 14 mGy and 21 mGy, respectively. AiCE provided consistently higher detection performance compared with AIDR-FC18 (48% higher for a 6 mm lumen diameter). Discrimination performance increased with increasing nominal diameter, dose, and for larger differences in stenosis severity. When comparing discrimination performance in HR to NR modes, the largest relative differences occur at the smallest nominal diameters and smallest differences in stenosis severity. The AiCE reconstruction algorithm produced the highest overall discrimination performance values, and these were significantly higher than AIDR-FC18 for nominal diameters of 7.14 and 10.08 mm.

CONCLUSIONS

HR mode outperforms NR for detection up to a specific diameter and the results improve with AiCE and for higher dose levels. For the task of size discrimination, HR mode consistently outperforms NR if AIDR-FC18 is used for dose levels of at least 21 mGy, and the results improve with AiCE and for the smallest differences in stenosis severity investigated (50% vs. 60%). High-resolution CT appears to be beneficial for detecting smaller simulated lumen diameters (<5 mm) and is generally advantageous for discrimination tasks related to stenotic lesions, which inherently contain information at higher frequencies, given the right reconstruction algorithm and dose level.

Combining Spatial Registration With Clutter Filtering for Power-Doppler Imaging in Peripheral Perfusion Applications

Power-Doppler ultrasonic (PD-US) imaging is sensitive to echoes from blood cell motion in the microvasculature but generally nonspecific because of difficulties with filtering nonblood-echo sources. We are studying the potential for using PD-US imaging for routine assessments of peripheral blood perfusion without contrast media. The strategy developed is based on an experimentally verified computational model of tissue perfusion that simulates typical in vivo conditions. The model considers directed and diffuse blood perfusion states in a field of moving clutter and noise. A spatial registration method is applied to minimize tissue motion prior to clutter and noise filtering. The results show that in-plane clutter motion is effectively minimized. While out-of-plane motion remains a strong source of clutter-filter leakage, those registration errors are readily minimized by straightforward modification of scanning techniques and spatial averaging.

Does scent attractiveness reveal women's ovulatory timing? Evidence from signal detection analyses and endocrine predictors of odour attractiveness

Odour cues associated with shifts in ovarian hormones indicate ovulatory timing in females of many nonhuman species. Although prior evidence supports women's body odours smelling more attractive on days when conception is possible, that research has left ambiguous how diagnostic of ovulatory timing odour cues are, as well as whether shifts in odour attractiveness are correlated with shifts in ovarian hormones. Here, 46 women each provided six overnight scent and corresponding day saliva samples spaced five days apart, and completed luteinizing hormone tests to determine ovulatory timing. Scent samples collected near ovulation were rated more attractive, on average, relative to samples from the same women collected on other days. Importantly, however, signal detection analyses showed that rater discrimination of fertile window timing from odour attractiveness ratings was very poor. Within-women shifts in salivary oestradiol and progesterone were not significantly associated with within-women shifts in odour attractiveness. Between-women, mean oestradiol was positively associated with mean odour attractiveness. Our findings suggest that raters cannot reliably detect women's ovulatory timing from their scent attractiveness. The between-women effect of oestradiol raises the possibility that women's scents provide information about overall cycle fecundity, though further research is necessary to rigorously investigate this possibility.

Assessment of Boundary Discrimination Performance in a Printed Phantom

Printed phantoms hold great potential as a tool for examining task-based image quality of x-ray imaging systems. Their ability to produce complex shapes rendered in materials with adjustable attenuation coefficients allows a new level of flexibility in the design of tasks for the evaluation of physical imaging systems. We investigate performance in a fine "boundary discrimination" task in which fine features at the margin of a clearly visible "lesion" are used to classify the lesion as malignant or benign. These tasks are appealing because of their relevance to clinical tasks, and because they typically emphasize higher spatial frequencies relative to more common lesion detection tasks. A 3D printed phantom containing cylindrical shells of varying thickness was used to generate lesions profiles that differed in their edge profiles. This was intended to approximate lesions with indistinct margins that are clinically associated with malignancy. Wall thickness in the phantom ranged from 0.4mm to 0.8mm, which allows for task difficulty to be varied by choosing different thicknesses to represent malignant and benign lesions. The phantom was immersed in a tub filled with water and potassium phosphate to approximate the attenuating background, and imaged repeatedly on a benchtop cone-beam CT scanner. After preparing the image data (reconstruction, ROI Selection, sub-pixel registration), we find that the mean frequency of the lesion profile is 0.11 cyc/mm. The mean frequency of the lesion-difference profile, representative of the discrimination task, is approximately 6 times larger. Model observers show appropriate dose performance in these tasks as well.

Computational Breast Anatomy Simulation Using Multi-Scale Perlin Noise

Virtual clinical trials (VCTs) of medical imaging require realistic models of human anatomy. For VCTs in breast imaging, a multi-scale Perlin noise method is proposed to simulate anatomical structures of breast tissue in the context of an ongoing breast phantom development effort. Four Perlin noise distributions were used to replace voxels representing the tissue compartments and Cooper's ligaments in the breast phantoms. Digital mammography and tomosynthesis projections were simulated using a clinical DBT system configuration. Power-spectrum analyses and higher-order statistics properties using Laplacian fractional entropy (LFE) of the parenchymal texture are presented. These objective measures were calculated in phantom and patient images using a sample of 140 clinical mammograms and 500 phantom images. Power-law exponents were calculated using the slope of the curve fitted in the low frequency [0.1, 1.0] mm-1 region of the power spectrum. The results show that the images simulated with our prior and proposed Perlin method have similar power-law spectra when compared with clinical mammograms. The power-law exponents calculated are -3.10, -3.55, and -3.46, for the log-power spectra of patient, prior phantom and proposed phantom images, respectively. The results also indicate an improved agreement between the mean LFE estimates of Perlin-noise based phantoms and patients than our prior phantoms and patients. Thus, the proposed method improved the simulation of anatomic noise substantially compared to our prior method, showing close agreement with breast parenchyma measures.

High-resolution μ CT imaging for characterizing microcalcification detection performance in breast CT

Purpose: To demonstrate the utility of high-resolution micro-computed tomography ( μ CT ) for determining ground-truth size and shape properties of calcium grains for evaluation of detection performance in breast CT (bCT). Approach: Calcium carbonate grains ( ∼ 200    μ m ) were suspended in 1% agar solution to emulate microcalcifications ( μ Calcs ) within a fibroglandular tissue background. Ground-truth imaging was performed on a commercial μ CT scanner and was used for assessing calcium-grain size and shape, and for generating μ Calc signal profiles. Calcium grains were placed within a realistic breast-shaped phantom and imaged on a prototype bCT system at 3- and 6-mGy mean glandular dose (MGD) levels, and the non-prewhitening detectability was assessed. Additionally, the μ CT -derived signal profiles were used in conjunction with the bCT system characterization (MTF and NPS) to obtain predictions of bCT detectability. Results: Estimated detectability of the calcium grains on the bCT system ranged from 2.5 to 10.6 for 3 mGy and from 3.8 to 15.3 for 6 mGy with large fractions of the grains meeting the Rose criterion for visibility. Segmentation of μ CT images based on morphological operations produced accurate results in terms of segmentation boundaries and segmented region size. A regression model linking bCT detectability to μ Calc parameters indicated significant effects of μ Calc size and vertical position within the breast phantom. Detectability using μ CT -derived detection templates and bCT statistical properties (MTF and NPS) were in good correspondence with those measured directly from bCT ( R 2 > 0.88 ). Conclusions: Parameters derived from μ CT ground-truth data were shown to produce useful characterizations of detectability when compared to estimates derived directly from bCT. Signal profiles derived from μ CT imaging can be used in conjunction with measured or hypothesized statistical properties to evaluate the performance of a system, or system component, that may not currently be available.

Special Section Guest Editorial: Conclusion to the Special Series on 2D and 3D Imaging: Perspectives in Human and Model Observer Performance

Guest editors Claudia Mello-Thoms, Craig K. Abbey, and Elizabeth A. Krupinski conclude the JMI Special Series on 2D and 3D Imaging, with commentary on the contributions.

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.

Comparative observer effects in 2D and 3D localization tasks

Purpose: Three-dimensional "volumetric" imaging methods are now a common component of medical imaging across many imaging modalities. Relatively little is known about how human observers localize targets masked by noise and clutter as they scroll through a 3D image and how it compares to a similar task confined to a single 2D slice. Approach: Gaussian random textures were used to represent noisy volumetric medical images. Subjects were able to freely inspect the images, including scrolling through 3D images as part of their search process. A total of eight experimental conditions were evaluated (2D versus 3D images, large versus small targets, power-law versus white noise). We analyze performance in these experiments using task efficiency and the classification image technique. Results: In 3D tasks, median response times were roughly nine times longer than 2D, with larger relative differences for incorrect trials. The efficiency data show a dissociation in which subjects perform with higher statistical efficiency in 2D tasks for large targets and higher efficiency in 3D tasks with small targets. The classification images suggest that a critical mechanism behind this dissociation is an inability to integrate across multiple slices to form a 3D localization response. The central slices of 3D classification images are remarkably similar to the corresponding 2D classification images. Conclusions: 2D and 3D tasks show similar weighting patterns between 2D images and the central slice of 3D images. There is relatively little weighting across slices in the 3D tasks, leading to lower task efficiency with respect to the ideal observer.

Under-exploration of Three-Dimensional Images Leads to Search Errors for Small Salient Targets

Advances in 3D imaging technology are transforming how radiologists search for cancer1,2 and how security officers scrutinize baggage for dangerous objects.3 These new 3D technologies often improve search over 2D images4,5 but vastly increase the image data. Here, we investigate 3D search for targets of various sizes in filtered noise and digital breast phantoms. For a Bayesian ideal observer optimally processing the filtered noise and a convolutional neural network processing the digital breast phantoms, search with 3D image stacks increases target information and improves accuracy over search with 2D images. In contrast, 3D search by humans leads to high miss rates for small targets easily detected in 2D search, but not for larger targets more visible in the visual periphery. Analyses of human eye movements, perceptual judgments, and a computational model with a foveated visual system suggest that human errors can be explained by interaction among a target's peripheral visibility, eye movement under-exploration of the 3D images, and a perceived overestimation of the explored area. Instructing observers to extend the search reduces 75% of the small target misses without increasing false positives. Results with twelve radiologists confirm that even medical professionals reading realistic breast phantoms have high miss rates for small targets in 3D search. Thus, under-exploration represents a fundamental limitation to the efficacy with which humans search in 3D image stacks and miss targets with these prevalent image technologies.

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.

High resolution microcalcification signal profiles for dedicated breast CT

This study introduces a methodology for generating high resolution signal profiles of microcalcification (MC) grains for validating breast CT (bCT) systems. A physical MC phantom was constructed by suspending calcium carbonate grains in an agar solution emulating MCs in a fibroglandular tissue background. Additionally, small Teflon spheres (2.4 mm diameter) were embedded in the agar solution for the purpose of fiducial marking and assessment of segmentation accuracy. The MC phantom was imaged on a high resolution (34 μm) commercial small-bore μCT scanner at high dose, and the images were used as the gold-standard for assessing MC size and for generating high resolution signal profiles of each MC. High-dose bCT scans of the MC phantom suspended in-air were acquired using 1 × 1 binning mode (75 μm dexel pitch) by averaging three repeat scans to produce a single low-noise reconstruction of the MC phantom. The high resolution μCT volume data set was then registered with the corresponding bCT data set after correcting for the bCT system spatial resolution. Microcalcification signal profiles constructed using low-noise bCT images were found to be in good agreement with those generated using the μCT scanner with all differences < 10% within the VOI surrounding each MC. The MC signal profiles were used as detection templates for a non-prewhitening-matched-filter model observer for scans acquired in a realistic breast phantom at 3, 6, and 9 mGy mean glandular dose. MC detectability using signal templates derived from bCT were shown to be in good agreement with those generated using μCT.

Introducing the Special Series on 2D and 3D Imaging: Perspectives in Human and Model Observer Performance

Guest editors Claudia Mello-Thoms, Craig Abbey, and Elizabeth A. Krupinski introduce the Special Series on 2D and 3D Imaging: Perspectives in Human and Model Observer Performance.

Effects of kV, filtration, dose, and object size on soft tissue and iodine contrast in dedicated breast CT

PURPOSE

Clinical use of dedicated breast computed tomography (bCT) requires relatively short scan times necessitating systems with high frame rates. This in turn impacts the x-ray tube operating range. We characterize the effects of tube voltage, beam filtration, dose, and object size on contrast and noise properties related to soft tissue and iodine contrast agents as a way to optimize imaging protocols for soft tissue and iodine contrast at high frame rates.

METHODS

This study design uses the signal-difference-to-noise ratio (SDNR), noise-equivalent quanta (NEQ), and detectability (d´) as measures of imaging performance for a prototype breast CT scanner that utilizes a pulsed x-ray tube (with a 4 ms pulse width) at 43.5 fps acquisition rate. We assess a range of kV, filtration, breast phantom size, and mean glandular dose (MGD). Performance measures are estimated from images of adipose-equivalent breast phantoms machined to have a representative size and shape of small, medium, and large breasts. Water (glandular tissue equivalent) and iodine contrast (5 mg/ml) were used to fill two cylindrical wells in the phantoms.

RESULTS

Air kerma levels required for obtaining an MGD of 6 mGy ranged from 7.1 to 9.1 mGy and are reported across all kV, filtration, and breast phantom sizes. However, at 50 kV, the thick filters (0.3 mm of Cu or Gd) exceeded the maximum available mA of the x-ray generator, and hence, these conditions were excluded from subsequent analysis. There was a strong positive association between measurements of SDNR and d' (R²  > 0.97) within the range of parameters investigated in this work. A significant decrease in soft tissue SDNR was observed for increasing phantom size and increasing kV with a maximum SDNR at 50 kV with 0.2 mm Cu or 0.2 mm Gd filtration. For iodine contrast SDNR, a significant decrease was observed with increasing phantom size, but a decrease in SDNR for increasing kV was only observed for 70 kV (50 and 60 kV were not significantly different). Thicker Gd filtration (0.3 mm Gd) resulted in a significant increase in iodine SDNR and decrease in soft tissue SDNR but requires significantly more tube current to deliver the same MGD.

CONCLUSIONS

The choice of 60 kV with 0.2 mm Gd filtration provides a good trade-off for maximizing both soft tissue and iodine contrast. This scanning technique takes advantage of the ~50 keV Gd k-edge to produce contrast and can be achieved within operating range of the x-ray generator used in this work. Imaging at 60 kV allows for a greater range in dose delivered to the large breast sizes when uniform image quality is desired across all breast sizes. While imaging performance metrics (i.e., detectability index and SDNR) were shown to be strongly correlated, the methodologies presented in this work for the estimation of NEQ (and subsequently d') provides a meaningful description of the spatial resolution and noise characteristics of this prototype bCT system across a range of beam quality, dose, and object sizes.

Human observer templates for lesion discrimination tasks

We investigate a series of two-alternative forced-choice (2AFC) discrimination tasks based on malignant features of abnormalities in low-dose lung CT scans. A total of 3 tasks are evaluated, and these consist of a size-discrimination task, a boundary-sharpness task, and an irregular-interior task. Target and alternative signal profiles for these tasks are modulated by one of two system transfer functions and embedded in ramp-spectrum noise that has been apodized for noise control in one of 4 different ways. This gives the resulting images statistical properties that are related to weak ground-glass lesions in axial slices of low-dose lung CT images. We investigate observer performance in these tasks using a combination of statistical efficiency and classification images. We report results of 24 2AFC experiments involving the three tasks. A staircase procedure is used to find the approximate 80% correct discrimination threshold in each task, with a subsequent set of 2,000 trials at this threshold. These data are used to estimate statistical efficiency with respect to the ideal observer for each task, and to estimate the observer template using the classification-image methodology. We find efficiency varies between the different tasks with lowest efficiency in the boundary-sharpness task, and highest efficiency in the non-uniform interior task. All three tasks produce clearly visible patterns of positive and negative weighting in the classification images. The spatial frequency plots of classification images show how apodization results in larger weights at higher spatial frequencies.

Deep Learning Based Model Observer by U-Net

Model Observers (MO) are algorithms designed to evaluate and optimize the parameters of new medical imaging reconstruction methodologies by providing a measure of human accuracy for a diagnostic task. In contrast with a computer-aided diagnosis system, MOs are not designed to outperform human diagnosis but only to find a defect if a radiologist would be able to detect it. These algorithms can economize and expedite the finding of optimal reconstruction parameters by reducing the number of sessions with expert radiologists, which are costly and prolonged. Convolutional Neural Networks (CNN or ConvNet) have been successfully used in the computer vision field for image classification, segmentation and video analytics. In this paper, we propose and test several U-Net configurations as MO for a defect localization task on synthetic images with different levels of correlated noisy backgrounds. Preliminary results show that the CNN based MO has potential and its accuracy correlates well with that of the human.

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.

Conspicuity of suspicious breast lesions on contrast enhanced breast CT compared to digital breast tomosynthesis and mammography

OBJECTIVE

Compare conspicuity of suspicious breast lesions on contrast-enhanced dedicated breast CT (CEbCT), tomosynthesis (DBT) and digital mammography (DM).

METHODS

100 females with BI-RADS 4/5 lesions underwent CEbCT and/or DBT prior to biopsy in this IRB approved, HIPAA compliant study. Two breast radiologists adjudicated lesion conspicuity scores (CS) for each modality independently. Data are shown as mean CS ±standard deviation. Two-sided t-test was used to determine significance between two modalities within each subgroup. Multiple comparisons were controlled by the false-discovery rate set to 5%.

RESULTS

50% of studied lesions were biopsy-confirmed malignancies. Malignant masses were more conspicuous on CEbCT than on DBT or DM (9.7 ±0.5, n = 25; 6.8 ± 3.1, n = 15; 6.7 ± 3.0, n = 27; p < 0.05). Malignant calcifications were equally conspicuous on all three modalities (CEbCT 8.7 ± 0.8, n = 18; DBT 8.5 ± 0.6, n = 15; DM 8.8 ± 0.7, n = 23; p = NS). Benign masses were equally conspicuous on CEbCT (6.6 ± 4.1, n = 22); DBT (6.4 ± 3.8, n = 17); DM (5.9 ± 3.6, n = 24; p = NS). Benign calcifications CS were similar between DBT (8.5 ± 1.0, n = 17) and DM (8.8 ± 0.8, n = 26; p = NS) but less conspicuous on CEbCT (4.0 ± 2.9, n = 25, p < 0.001). 55 females were imaged with all modalities. Results paralleled the entire cohort. 69%(n = 62) of females imaged by CEbCT had dense breasts. Benign/malignant lesion CSs in dense/non-dense categories were 4.8 ± 3.7, n = 33, vs 6.0 ± 3.9, n = 14, p = 0.35; 9.2 ± 0.9, n = 29 vs. 9.4 ± 0.7, n = 14; p = 0.29, respectively.

CONCLUSION

Malignant masses are more conspicuous on CEbCT than DM or DBT. Malignant microcalcifications are equally conspicuous on all three modalities. Benign calcifications remain better visualized by DM and DBT than with CEbCT. We observed no differences in benign masses on all modalities. CS of both benign and malignant lesions were independent of breast density.

ADVANCES IN KNOWLEDGE

CEbCT is a promising diagnostic imaging modality for suspicious breast lesions.

Evaluation of non-Gaussian statistical properties in virtual breast phantoms

Images derived from a "virtual phantom" can be useful in characterizing the performance of imaging systems. This has driven the development of virtual breast phantoms implemented in simulation environments. In breast imaging, several such phantoms have been proposed. We analyze the non-Gaussian statistical properties from three classes of virtual breast phantoms and compare them to similar statistics from a database of breast images. These include clustered-blob lumpy backgrounds (CBLBs), truncated binary textures, and the UPenn virtual breast phantoms. We use Laplacian fractional entropy (LFE) as a measure of the non-Gaussian statistical properties of each simulation procedure. Our results show that, despite similar power spectra, the simulation approaches differ considerably in LFE with very low scores for the CBLB to high values for the UPenn phantom at certain frequencies. These results suggest that LFE may have value in developing and tuning virtual phantom simulation procedures.

Method for Recovering Lost Ultrasonic Information Using the Echo-intensity Mean

The axial resolution of a B-mode (or intensity) image is limited by the bandwidth of the pulse envelope. In this report, we investigate the source of this limitation by examining the transfer of high-resolution information from the tissue impedance variance throughout the imaging process. For that purpose, we express the mean and variance of the echo-intensity signal as a linear system to track the flow of object information along the image-formation chain. The results reveal how demodulation influences the available information by discarding high spatial-frequency information. This analysis further points to a simple way to recover lost information with only a minor addition to signal processing. Software phantoms are used to show that under ideal conditions, information from small-scale high-contrast reflectors, such as microcalcifications, can be significantly enhanced with this simple change to echo processing.

Inter-laboratory comparison of channelized hotelling observer computation

PURPOSE

The task-based assessment of image quality using model observers is increasingly used for the assessment of different imaging modalities. However, the performance computation of model observers needs standardization as well as a well-established trust in its implementation methodology and uncertainty estimation. The purpose of this work was to determine the degree of equivalence of the channelized Hotelling observer performance and uncertainty estimation using an intercomparison exercise.

MATERIALS AND METHODS

Image samples to estimate model observer performance for detection tasks were generated from two-dimensional CT image slices of a uniform water phantom. A common set of images was sent to participating laboratories to perform and document the following tasks: (a) estimate the detectability index of a well-defined CHO and its uncertainty in three conditions involving different sized targets all at the same dose, and (b) apply this CHO to an image set where ground truth was unknown to participants (lower image dose). In addition, and on an optional basis, we asked the participating laboratories to (c) estimate the performance of real human observers from a psychophysical experiment of their choice. Each of the 13 participating laboratories was confidentially assigned a participant number and image sets could be downloaded through a secure server. Results were distributed with each participant recognizable by its number and then each laboratory was able to modify their results with justification as model observer calculation are not yet a routine and potentially error prone.

RESULTS

Detectability index increased with signal size for all participants and was very consistent for 6 mm sized target while showing higher variability for 8 and 10 mm sized target. There was one order of magnitude between the lowest and the largest uncertainty estimation.

CONCLUSIONS

This intercomparison helped define the state of the art of model observer performance computation and with thirteen participants, reflects openness and trust within the medical imaging community. The performance of a CHO with explicitly defined channels and a relatively large number of test images was consistently estimated by all participants. In contrast, the paper demonstrates that there is no agreement on estimating the variance of detectability in the training and testing setting.

Classification images for localization performance in ramp-spectrum noise

PURPOSE

This study investigates forced localization of targets in simulated images with statistical properties similar to trans-axial sections of x-ray computed tomography (CT) volumes. A total of 24 imaging conditions are considered, comprising two target sizes, three levels of background variability, and four levels of frequency apodization. The goal of the study is to better understand how human observers perform forced-localization tasks in images with CT-like statistical properties.

METHODS

The transfer properties of CT systems are modeled by a shift-invariant transfer function in addition to apodization filters that modulate high spatial frequencies. The images contain noise that is the combination of a ramp-spectrum component, simulating the effect of acquisition noise in CT, and a power-law component, simulating the effect of normal anatomy in the background, which are modulated by the apodization filter as well. Observer performance is characterized using two psychophysical techniques: efficiency analysis and classification image analysis. Observer efficiency quantifies how much diagnostic information is being used by observers to perform a task, and classification images show how that information is being accessed in the form of a perceptual filter.

RESULTS

Psychophysical studies from five subjects form the basis of the results. Observer efficiency ranges from 29% to 77% across the different conditions. The lowest efficiency is observed in conditions with uniform backgrounds, where significant effects of apodization are found. The classification images, estimated using smoothing windows, suggest that human observers use center-surround filters to perform the task, and these are subjected to a number of subsequent analyses. When implemented as a scanning linear filter, the classification images appear to capture most of the observer variability in efficiency (r² = 0.86). The frequency spectra of the classification images show that frequency weights generally appear bandpass in nature, with peak frequency and bandwidth that vary with statistical properties of the images.

CONCLUSIONS

In these experiments, the classification images appear to capture important features of human-observer performance. Frequency apodization only appears to have a significant effect on performance in the absence of anatomical variability, where the observers appear to underweight low spatial frequencies that have relatively little noise. Frequency weights derived from the classification images generally have a bandpass structure, with adaptation to different conditions seen in the peak frequency and bandwidth. The classification image spectra show relatively modest changes in response to different levels of apodization, with some evidence that observers are attempting to rebalance the apodized spectrum presented to them.

Evaluation of Search Strategies for Microcalcifications and Masses in 3D Images

Medical imaging is quickly evolving towards 3D image modalities such as computed tomography (CT), magnetic resonance imaging (MRI) and digital breast tomosynthesis (DBT). These 3D image modalities add volumetric information but further increase the need for radiologists to search through the image data set. Although much is known about search strategies in 2D images less is known about the functional consequences of different 3D search strategies. We instructed readers to use two different search strategies: drillers had their eye movements restricted to a few regions while they quickly scrolled through the image stack, scanners explored through eye movements the 2D slices. We used real-time eye position monitoring to ensure observers followed the drilling or the scanning strategy while approximately preserving the percentage of the volumetric data covered by the useful field of view. We investigated search for two signals: a simulated microcalcification and a larger simulated mass. Results show an interaction between the search strategy and lesion type. In particular, scanning provided significantly better detectability for microcalcifications at the cost of 5 times more time to search while there was little change in the detectability for the larger simulated masses. Analyses of eye movements support the hypothesis that the effectiveness of a search strategy in 3D imaging arises from the interaction of the fixational sampling of visual information and the signals' visibility in the visual periphery.

Interactions of lesion detectability and size across single-slice DBT and 3D DBT

Three dimensional image modalities introduce a new paradigm for visual search requiring visual exploration of a larger search space than 2D imaging modalities. The large number of slices in the 3D volumes and the limited reading times make it difficult for radiologists to explore thoroughly by fixating with their high resolution fovea on all regions of each slice. Thus, for 3D images, observers must rely much more on their visual periphery (points away from fixation) to process image information. We previously found a dissociation in signal detectability between 2D and 3D search tasks for small signals in synthetic textures evaluated with non-radiologist trained observers. Here, we extend our evaluation to more clinically realistic backgrounds and radiologist observers. We studied the detectability of simulated microcalcifications (MCALC) and masses (MASS) in Digital Breast Tomosynthesis (DBT) utilizing virtual breast phantoms. We compared the lesion detectability of 8 radiologists during free search in 3D DBT and a 2D single-slice DBT (center slice of the 3D DBT). Our results show that the detectability of the microcalcification degrades significantly in 3D DBT with respect to the 2D single-slice DBT. On the other hand, the detectability for masses does not show this behavior and its detectability is not significantly different. The large deterioration of the 3D detectability of microcalcifications relative to masses may be related to the peripheral processing given the high number of cases in which the microcalcification was missed and the high number of search errors. Together, the results extend previous findings with synthetic textures and highlight how search in 3D images is distinct from 2D search as a consequence of the interaction between search strategies and the visibility of signals in the visual periphery.

Visual adaptation and the amplitude spectra of radiological images

We examined how visual sensitivity and perception are affected by adaptation to the characteristic amplitude spectra of X-ray mammography images. Because of the transmissive nature of X-ray photons, these images have relatively more low-frequency variability than natural images, a difference that is captured by a steeper slope of the amplitude spectrum (~ − 1.5) compared to the ~ 1/f (slope of − 1) spectra common to natural scenes. Radiologists inspecting these images are therefore exposed to a different balance of spectral components, and we measured how this exposure might alter spatial vision. Observers (who were not radiologists) were adapted to images of normal mammograms or the same images sharpened by filtering the amplitude spectra to shallower slopes. Prior adaptation to the original mammograms significantly biased judgments of image focus relative to the sharpened images, demonstrating that the images are sufficient to induce substantial after-effects. The adaptation also induced strong losses in threshold contrast sensitivity that were selective for lower spatial frequencies, though these losses were very similar to the threshold changes induced by the sharpened images. Visual search for targets (Gaussian blobs) added to the images was also not differentially affected by adaptation to the original or sharper images. These results complement our previous studies examining how observers adapt to the textural properties or phase spectra of mammograms. Like the phase spectrum, adaptation to the amplitude spectrum of mammograms alters spatial sensitivity and visual judgments about the images. However, unlike the phase spectrum, adaptation to the amplitude spectra did not confer a selective performance advantage relative to more natural spectra.

Expanding Acquisition and Clutter Filter Dimensions for Improved Perfusion Sensitivity

A method is explored for increasing the sensitivity of power-Doppler imaging without contrast enhancement. We acquire 1-10 s of echo signals and arrange it into a 3-D spatiotemporal data array. An eigenfilter developed to preserve all three dimensions of the array yields power estimates for blood flow and perfusion that are well separated from tissue clutter. This method is applied at high frequency (24-MHz pulses) to a murine model of an ischemic hindlimb. We demonstrate enhancements to tissue perfusion maps in normal and ischemic tissues. The method can be applied to data from any ultrasonic instrument that provides beamformed RF echo data.

Linear System Models for Ultrasonic Imaging: Intensity Signal Statistics

Despite a great deal of work characterizing the statistical properties of radio frequency backscattered ultrasound signals, less is known about the statistical properties of demodulated intensity signals. Analysis of intensity is made more difficult by a strong nonlinearity that arises in the process of demodulation. This limits our ability to characterize the spatial resolution and noise properties of B-mode ultrasound images. In this paper, we generalize earlier results on two-point intensity covariance using a multivariate systems approach. We derive the mean and autocovariance function of the intensity signal under Gaussian assumptions on both the object scattering function and acquisition noise, and with the assumption of a locally shift-invariant pulse-echo system function. We investigate the limiting cases of point statistics and a uniform scattering field with a stationary distribution. Results from validation studies using simulation and data from a real system applied to a uniform scattering phantom are presented. In the simulation studies, we find errors less than 10% between the theoretical mean and variance, and sample estimates of these quantities. Prediction of the intensity power spectrum (PS) in the real system exhibits good qualitative agreement (errors less than 3.5 dB for frequencies between 0.1 and 10 cyc/mm, but with somewhat higher error outside this range that may be due to the use of a window in the PS estimation procedure). We also replicate the common finding that the intensity mean is equal to its standard deviation (i.e., signal-to-noise ratio = 1) for fully developed speckle. We show how the derived statistical properties can be used to characterize the quality of an ultrasound linear array for low-contrast patterns using generalized noise-equivalent quanta directly on the intensity signal.

The role of extra-foveal processing in 3D imaging

The field of medical image quality has relied on the assumption that metrics of image quality for simple visual detection tasks are a reliable proxy for the more clinically realistic visual search tasks. Rank order of signal detectability across conditions often generalizes from detection to search tasks. Here, we argue that search in 3D images represents a paradigm shift in medical imaging: radiologists typically cannot exhaustively scrutinize all regions of interest with the high acuity fovea requiring detection of signals with extra-foveal areas (visual periphery) of the human retina. We hypothesize that extra-foveal processing can alter the detectability of certain types of signals in medical images with important implications for search in 3D medical images. We compare visual search of two different types of signals in 2D vs. 3D images. We show that a small microcalcification-like signal is more highly detectable than a larger mass-like signal in 2D search, but its detectability largely decreases (relative to the larger signal) in the 3D search task. Utilizing measurements of observer detectability as a function retinal eccentricity and observer eye fixations we can predict the pattern of results in the 2D and 3D search studies. Our findings: 1) suggest that observer performance findings with 2D search might not always generalize to 3D search; 2) motivate the development of a new family of model observers that take into account the inhomogeneous visual processing across the retina (foveated model observers).

Foveated Model Observers to predict human performance in 3D images

We evaluate 3D search requires model observers that take into account the peripheral human visual processing (foveated models) to predict human observer performance. We show that two different 3D tasks, free search and location-known detection, influence the relative human visual detectability of two signals of different sizes in synthetic backgrounds mimicking the noise found in 3D digital breast tomosynthesis. One of the signals resembled a microcalcification (a small and bright sphere), while the other one was designed to look like a mass (a larger Gaussian blob). We evaluated current standard models observers (Hotelling; Channelized Hotelling; non-prewhitening matched filter with eye filter, NPWE; and non-prewhitening matched filter model, NPW) and showed that they incorrectly predict the relative detectability of the two signals in 3D search. We propose a new model observer (3D Foveated Channelized Hotelling Observer) that incorporates the properties of the visual system over a large visual field (fovea and periphery). We show that the foveated model observer can accurately predict the rank order of detectability of the signals in 3D images for each task. Together, these results motivate the use of a new generation of foveated model observers for predicting image quality for search tasks in 3D imaging modalities such as digital breast tomosynthesis or computed tomography.

Optimal Joint Detection and Estimation That Maximizes ROC-Type Curves

Combined detection-estimation tasks are frequently encountered in medical imaging. Optimal methods for joint detection and estimation are of interest because they provide upper bounds on observer performance, and can potentially be utilized for imaging system optimization, evaluation of observer efficiency, and development of image formation algorithms. We present a unified Bayesian framework for decision rules that maximize receiver operating characteristic (ROC)-type summary curves, including ROC, localization ROC (LROC), estimation ROC (EROC), free-response ROC (FROC), alternative free-response ROC (AFROC), and exponentially-transformed FROC (EFROC) curves, succinctly summarizing previous results. The approach relies on an interpretation of ROC-type summary curves as plots of an expected utility versus an expected disutility (or penalty) for signal-present decisions. We propose a general utility structure that is flexible enough to encompass many ROC variants and yet sufficiently constrained to allow derivation of a linear expected utility equation that is similar to that for simple binary detection. We illustrate our theory with an example comparing decision strategies for joint detection-estimation of a known signal with unknown amplitude. In addition, building on insights from our utility framework, we propose new ROC-type summary curves and associated optimal decision rules for joint detection-estimation tasks with an unknown, potentially-multiple, number of signals in each observation.

Efficiency of U.S. Tissue Perfusion Estimators

We measure the detection and discrimination efficiencies of conventional power-Doppler estimation of perfusion without contrast enhancement. The measurements are made in a phantom with known blood-mimicking fluid flow rates in the presence of clutter and noise. Efficiency is measured by comparing functions of the areas under the receiver operating characteristic curve for Doppler estimators with those of the ideal discriminator, for which we estimate the temporal covariance matrix from echo data. Principal-component analysis is examined as a technique for increasing the accuracy of covariance matrices estimated from echo data. We find that Doppler estimators are <50% efficient at directed perfusion detection between 0.1 and 2.0 mL/min per 2 cm(2) flow area. The efficiency was 20%-40% for the task of discriminating between two perfusion rates in the same range. We conclude that there are reasons to search for more efficient perfusion estimators, one that incorporates covariance matrix information that could significantly enhance the utility of Doppler ultrasound without contrast enhancement.

Discriminatory power of common genetic variants in personalized breast cancer diagnosis

Technology advances in genome-wide association studies (GWAS) has engendered optimism that we have entered a new age of precision medicine, in which the risk of breast cancer can be predicted on the basis of a person's genetic variants. The goal of this study is to evaluate the discriminatory power of common genetic variants in breast cancer risk estimation. We conducted a retrospective case-control study drawing from an existing personalized medicine data repository. We collected variables that predict breast cancer risk: 153 high-frequency/low-penetrance genetic variants, reflecting the state-of-the-art GWAS on breast cancer, mammography descriptors and BI-RADS assessment categories in the Breast Imaging Reporting and Data System (BI-RADS) lexicon. We trained and tested naïve Bayes models by using these predictive variables. We generated ROC curves and used the area under the ROC curve (AUC) to quantify predictive performance. We found that genetic variants achieved comparable predictive performance to BI-RADS assessment categories in terms of AUC (0.650 vs. 0.659, p-value = 0.742), but significantly lower predictive performance than the combination of BI-RADS assessment categories and mammography descriptors (0.650 vs. 0.751, p-value < 0.001). A better understanding of relative predictive capability of genetic variants and mammography data may benefit clinicians and patients to make appropriate decisions about breast cancer screening, prevention, and treatment in the era of precision medicine.

A Utility/Cost Analysis of Breast Cancer Risk Prediction Algorithms

Breast cancer risk prediction algorithms are used to identify subpopulations that are at increased risk for developing breast cancer. They can be based on many different sources of data such as demographics, relatives with cancer, gene expression, and various phenotypic features such as breast density. Women who are identified as high risk may undergo a more extensive (and expensive) screening process that includes MRI or ultrasound imaging in addition to the standard full-field digital mammography (FFDM) exam. Given that there are many ways that risk prediction may be accomplished, it is of interest to evaluate them in terms of expected cost, which includes the costs of diagnostic outcomes. In this work we perform an expected-cost analysis of risk prediction algorithms that is based on a published model that includes the costs associated with diagnostic outcomes (true-positive, false-positive, etc.). We assume the existence of a standard screening method and an enhanced screening method with higher scan cost, higher sensitivity, and lower specificity. We then assess expected cost of using a risk prediction algorithm to determine who gets the enhanced screening method under the strong assumption that risk and diagnostic performance are independent. We find that if risk prediction leads to a high enough positive predictive value, it will be cost-effective regardless of the size of the subpopulation. Furthermore, in terms of the hit-rate and false-alarm rate of the of the risk-prediction algorithm, iso-cost contours are lines with slope determined by properties of the available diagnostic systems for screening.

Differentiation of ductal carcinoma in-situ from benign micro-calcifications by dedicated breast computed tomography

PURPOSE

Compare conspicuity of ductal carcinoma in-situ (DCIS) to benign calcifications on unenhanced (bCT), contrast-enhanced dedicated breast CT (CEbCT) and mammography (DM).

METHODS AND MATERIALS

The institutional review board approved this HIPAA-compliant study. 42 women with Breast Imaging Reporting and Data System 4 or 5 category micro-calcifications had breast CT before biopsy. Three subjects with invasive disease at surgery were excluded. Two breast radiologists independently compared lesion conspicuity scores (CS) for CEbCT, to bCT and DM. Enhancement was measured in Hounsfield units (HU). Mean CS ± standard deviations are shown. Receiver operating characteristic analysis (ROC) measured radiologists' discrimination performance by comparing CS to enhancement alone. Statistical measurements were made using ANOVA F-test, Wilcoxon rank-sum test and robust linear regression analyses.

RESULTS

39 lesions (17 DCIS, 22 benign) were analyzed. DCIS (8.5 ± 0.9, n=17) was more conspicuous than benign micro-calcifications (3.6 ± 2.9, n=22; p<0.0001) on CEbCT. DCIS was equally conspicuous on CEbCT and DM (8.5 ± 0.9, 8.7 ± 0.8, n=17; p=0.85) and more conspicuous when compared to bCT (5.3 ± 2.6, n=17; p<0.001). All DCIS enhanced; mean enhancement (90HU ± 53HU, n=17) was higher compared to benign lesions (33 ± 30HU, n=22) (p<0.0001). ROC analysis of the radiologists' CS showed high discrimination performance (AUC=0.94) compared to enhancement alone (AUC=0.85) (p<0.026).

CONCLUSION

DCIS is more conspicuous than benign micro-calcifications on CEbCT. DCIS visualization on CEbCT is equal to mammography but improved compared to bCT. Radiologists' discrimination performance using CEBCT is significantly higher than enhancement values alone. CEbCT may have an advantage over mammography by reducing false positive examinations when calcifications are analyzed.

Developing a utility decision framework to evaluate predictive models in breast cancer risk estimation

Combining imaging and genetic information to predict disease presence and progression is being codified into an emerging discipline called "radiogenomics." Optimal evaluation methodologies for radiogenomics have not been well established. We aim to develop a decision framework based on utility analysis to assess predictive models for breast cancer diagnosis. We garnered Gail risk factors, single nucleotide polymorphisms (SNPs), and mammographic features from a retrospective case-control study. We constructed three logistic regression models built on different sets of predictive features: (1) Gail, (2) Gail + Mammo, and (3) Gail + Mammo + SNP. Then we generated receiver operating characteristic (ROC) curves for three models. After we assigned utility values for each category of outcomes (true negatives, false positives, false negatives, and true positives), we pursued optimal operating points on ROC curves to achieve maximum expected utility of breast cancer diagnosis. We performed McNemar's test based on threshold levels at optimal operating points, and found that SNPs and mammographic features played a significant role in breast cancer risk estimation. Our study comprising utility analysis and McNemar's test provides a decision framework to evaluate predictive models in breast cancer risk estimation.