Showing papers by "Craig K. Abbey published in 2010"

An Equivalent Relative Utility Metric for Evaluating Screening Mammography

Abstract: Comparative studies of performance in screening mammography are often ambiguous. A new method will frequently show a higher sensitivity or detection rate than an existing standard with a concomitan...

Optimal beamforming in ultrasound using the ideal observer

Abstract: Beamforming of received pulse-echo data generally involves the compression of signals from multiple channels within an aperture. This compression is irreversible, and therefore allows the possibility that information relevant for performing a diagnostic task is irretrievably lost. The purpose of this study was to evaluate information transfer in beamforming using a previously developed ideal observer model to quantify diagnostic information relevant to performing a task. We describe an elaborated statistical model of image formation for fixed-focus transmission and single-channel reception within a moving aperture, and we use this model on a panel of tasks related to breast sonography to evaluate receive-beamforming approaches that optimize the transfer of information. Under the assumption that acquisition noise is well described as an additive wide-band Gaussian white-noise process, we show that signal compression across receive-aperture channels after a 2-D matched-filtering operation results in no loss of diagnostic information. Across tasks, the matched-filter beamformer results in more information than standard delay-and-sum beamforming in the subsequent radio-frequency signal by a factor of two. We also show that for this matched filter, 68% of the information gain can be attributed to the phase of the matched-filter and 21% can be attributed to the amplitude. A 1-D matched filtering along axial lines shows no advantage over delay-andsum, suggesting an important role for incorporating correlations across different aperture windows in beamforming. We also show that a post-compression processing before the computation of an envelope is necessary to pass the diagnostic information in the beamformed radio-frequency signal to the final envelope image.

Model Observers for Complex Discrimination Tasks: Assessments of Multiple Coronary Stent Placements

Abstract: As an important clinical task, evaluating the placement of multiple coronary stents requires fine judgments of distance between stents. However, making these judgments is limited by low system resolution, noise, low contrast of the deployed stent, and stent motion during the cardiac cycle. We use task performance as a figure of merit for optimizing image display parameters. In previous work, we described our simulation procedure in detail, and also reported results of human observers for a visual task involving discrimination of 4 gap sizes under various frame rates and number of frames. Here, we report the results of three spatial model observers (i.e. NPW, NPWE, and PWMF) and two temporal sensitivity functions (i.e. transient and sustained) for the same task. Under signal known exactly conditions, we find that model observers can be used to predict human observers in terms of discrimination accuracy by adding internal noise.

High human-observer efficiency for forced-localization tasks in correlated noise

Abstract: We investigate signal-known-exactly (SKE) detection and free localization tasks in correlated noise having a power-law power spectrum. In all cases the target is an additive focal Gaussian "bump" signal. We compare the performance of human observers in psychophysical studies to that of the Bayesian ideal observer for both tasks in the form of the observer efficiency. We find efficiencies that range from 40% to 60% in the SKE detection task, consistent with previous works. Observer efficiency is considerably higher in the free-localization task, ranging from 60% to 80%. Direct estimation of the human-observer spatial-frequency weights shows clear evidence of a shift to higher frequencies with increasing power-law exponent in both tasks. Our results suggest that human observers are performing some sort of limited prewhitening of stimuli in both tasks.

Tomographic reconstruction of the pulse-echo spatiotemporal impulse response

Abstract: Virtually every area of ultrasonic imaging research requires accurate estimation of the spatiotemporal impulse response of the instrument, and yet accurate measurements are difficult to achieve. The impulse response can also be difficult to predict numerically for a specific device because small unknown perturbations in array properties can generate significant changes in predicted pulse-echo field patterns. A typical measurement for a 1-D array transducer employs a line scatterer oriented perpendicular to the scan plane. Echoes from line scatterers located throughout the field of view constitute estimates of shift-varying line response functions. We propose an inverse-problem approach to the reconstruction of point-spread functions from line-spread functions. A collection of echoes recorded for a range of line-scatterer rotation angles are treated as projections of sound pressure onto the transducer array surface. Although the reconstruction is mathematically equivalent to filtered backprojection, it provides significant advantages with respect to interpolation that confound straightforward implementations. Field II predictions used to model measurements made on commercial systems suggest the reconstruction accuracy is with 0.32% for noiseless echo data. Application of the method to data acquired from a commercial system are evaluated from the perspective of deconvolution.

Ideal observer comparison between tomographic and projection x-ray images of the breast

Abstract: The image data from 348 breast CT studies performed on patient volunteers in a Phase II clinical trial were used in a pre-whitened matched filter (PWMF) model observer study This computer-based investigation simulated the addition of spherical lesions of various diameters to the breast data sets, and a series of images with different slice thicknesses ranging from 0.3 mm to 44 mm were created by voxel averaging The PWMF was tailored to each slice thickness, and signal known exactly receiver operating characteristic analysis was performed A total of 1000 lesions and “non-lesions” were simulated for each diameter on each breast data set Receiver operating characteristic (ROC) curve analysis was performed, and sensitivity at 95% specificity was computed Thin slice imaging as with breast CT demonstrated significantly better sensitivity than thick slice images as with mammography, with improvements in sensitivity ranging from 5% to 50%.

Robustness of Wiener filtering as an approximation to the Bayesian observer stragegy

Abstract: Wiener filters applied to channel-summed RF echo data was approximated to the ideal Bayesian beamforming strategy for imaging breast lesion features They can be computed in real-time and perform well provided the system response matrix is circulant Real systems exhibit depth-varying impulse responses, so performance degrades for some essential diagnostic features This paper describes the effects of circulant assumption (CA) violations on the robustness of human visual discrimination efficiency, and the effects of regularization, as applied to beamformers for breast sonography The Wiener filter is then adapted for shift-varying system responses The trade off between performance and computation time for each approach is studied using simulations and phantoms scanned with a commercial system

Cystic resolution and task performance in beamforming

Abstract: Cystic resolution (CR) may be considered to be the contrast resolution in the center of a small anechoic “cyst”. The goal of this work is to evaluate cystic resolution from the perspective of task performance in breast sonography. We demonstrate that CR is rigorously related to task performance for the particular task of detecting the presence of a cyst from a single sample at the center of the cyst. We compare this to other tasks relevant to breast imaging.