Detection performance theory for ultrasound imaging systems
TL;DR: Close-form and low-contrast approximations of ideal observer performance are derived for signal known statistically detection tasks and a metric borrowed and modified from photon imaging, Generalized Noise Equivalent Quanta, is shown to be a useful and measurable target-independent figure of merit when adapted for ultrasound systems.
Abstract: A rigorous statistical theory for characterizing the performance of medical ultrasound systems for lesion detection tasks is developed. A design strategy for optimizing ultrasound systems should be to adjust parameters for maximum information content, which is obtained by maximizing the ideal observer performance. Then, given the radio-frequency data, image and signal processing algorithms are designed to extract as much diagnostically relevant information as possible. In this paper, closed-form and low-contrast approximations of ideal observer performance are derived for signal known statistically detection tasks. The accuracy of the approximations are tested by comparing with Monte Carlo techniques. A metric borrowed and modified from photon imaging, Generalized Noise Equivalent Quanta, is shown to be a useful and measurable target-independent figure of merit when adapted for ultrasound systems. This theory provides the potential to optimize design tradeoffs for detection tasks. For example it may help us understand how to push the limits of specific features, such as spatial resolution, without significantly compromising overall detection performance.
Citations
More filters
TL;DR: A model-based signal decomposition scheme is proposed that identifies spatial frequency signatures to decompose received wavefronts into their most significant scattering sources and shows the decluttered B-mode images have an average improvement in contrast over normal images.
Abstract: In vivo ultrasonic imaging with transducer arrays suffers from image degradation resulting from beamforming limitations, including diffraction-limited beamforming and beamforming degradation caused by tissue inhomogeneity. Additionally, based on recent studies, multipath scattering also causes significant image degradation. To reduce degradation from both sources, we propose a model-based signal decomposition scheme. The proposed algorithm identifies spatial frequency signatures to decompose received wavefronts into their most significant scattering sources. Scattering sources originating from a region of interest are used to reconstruct decluttered wavefronts, which are beamformed into decluttered RF scan lines or A-lines. To test the algorithm, ultrasound system channel data were acquired during liver scans from 8 patients. Multiple data sets were acquired from each patient, with 55 total data sets, 43 of which had identifiable hypoechoic regions on normal B-mode images. The data sets with identifiable hypoechoic regions were analyzed. The results show the decluttered B-mode images have an average improvement in contrast over normal images of 7.3 ± 4.6 dB. The contrast-to-noise ratio (CNR) changed little on average between normal and decluttered Bmode, -0.4 ± 5.9 dB. The in vivo speckle SNR decreased; the change was -0.65 ± 0.28. Phantom speckle SNR also decreased, but only by -0.40 ± 0.03.
69 citations
Cites background from "Detection performance theory for ul..."
...Generally, in recent literature, beamforming based image degradation—including diffraction limitations and sound speed inhomogeneity—is widely considered to be the dominant source of in vivo image corruption [4]–[7]....
[...]
TL;DR: A numerical approach for evaluating the ideal observer acting on radio frequency (RF) frame data, which involves inversion of large nonstationary covariance matrices, and a power-series approach to computing this inverse is described.
Abstract: We investigate and extend the ideal observer methodology developed by Smith and Wagner to detection and discrimination tasks related to breast sonography. We provide a numerical approach for evaluating the ideal observer acting on radio frequency (RF) frame data, which involves inversion of large nonstationary covariance matrices, and we describe a power-series approach to computing this inverse. Considering a truncated power series suggests that the RF data be Wiener-filtered before forming the final envelope image. We have compared human performance for Wiener-filtered and conventional B-mode envelope images using psychophysical studies for 5 tasks related to breast cancer classification. We find significant improvements in visual detection and discrimination efficiency in four of these five tasks. We also use the Smith-Wagner approach to distinguish between human and processing inefficiencies, and find that generally the principle limitation comes from the information lost in computing the final envelope image.
64 citations
TL;DR: This work derives decision strategies that maximize the area under the LROC curve, A/sub LROC/.
Abstract: For the 2-class detection problem (signal absent/present), the likelihood ratio is an ideal observer in that it minimizes Bayes risk for arbitrary costs and it maximizes the area under the receiver operating characteristic (ROC) curve [AUC]. The AUC-optimizing property makes it a valuable tool in imaging system optimization. If one considered a different task, namely, joint detection and localization of the signal, then it would be similarly valuable to have a decision strategy that optimized a relevant scalar figure of merit. We are interested in quantifying performance on decision tasks involving location uncertainty using the localization ROC (LROC) methodology. Therefore, we derive decision strategies that maximize the area under the LROC curve, A/sub LROC/. We show that these decision strategies minimize Bayes risk under certain reasonable cost constraints. The detection-localization task is modeled as a decision problem in three increasingly realistic ways. In the first two models, we treat location as a discrete parameter having finitely many values resulting in an (L+1) class classification problem. In our first simple model, we do not include search tolerance effects and in the second, more general, model, we do. In the third and most general model, we treat location as a continuous parameter and also include search tolerance effects. In all cases, the essential proof that the observer maximizes A/sub LROC/ is obtained with a modified version of the Neyman-Pearson lemma. A separate form of proof is used to show that in all three cases, the decision strategy minimizes the Bayes risk under certain reasonable cost constraints.
55 citations
Cites background from "Detection performance theory for ul..."
...imaging system optimization [2]–[4] and other applications [5]....
[...]
TL;DR: Analytical beam pattern analysis and simulation results show that this spatial filtering method can improve lateral resolution and contrast-to-noise ratio as compared with conventional dynamic receive focusing (DRF) methods.
Abstract: This paper proposes an efficient array beam-forming method using spatial matched filtering (SMF) for ultrasonic imaging. In the proposed method, ultrasonic waves are transmitted from an array subaperture with fixed transmit focus as in conventional array imaging. At receive, radio frequency echo signals from each receive channel are passed through a spatial matched filter that is constructed based on the system transmit-receive spatial impulse response. The filtered echo signals are then summed without time delays. The filter concentrates and spatially registers the echo energy from each element so that the pulse-echo impulse response of the summed output is focused with acceptably low side lobes. Analytical beam pattern analysis and simulation results using a linear array show that this spatial filtering method can improve lateral resolution and contrast-to-noise ratio as compared with conventional dynamic receive focusing (DRF) methods. Experimental results with a linear array are consistent but point out the need to address additional practical issues. Spatial filtering is equivalent to synthetic aperture methods that dynamically focus on both transmit and receive throughout the field of view. In one common example of phase aberrations, the SMF method was degraded to a degree comparable to conventional DRF methods.
45 citations
TL;DR: This work proposes the use of histogram matching to better assess differences across image formation methods and presents variations of histograms matching and provides code to encourage the application of this method within the imaging community.
Abstract: The widespread development of new ultrasound image formation techniques has created a need for a standardized methodology for comparing the resulting images. Traditional methods of evaluation use quantitative metrics to assess the imaging performance in specific tasks, such as point resolution or lesion detection. Quantitative evaluation is complicated by unconventional new methods and nonlinear transformations of the dynamic range of data and images. Transformation-independent image metrics have been proposed for quantifying task performance. However, clinical ultrasound still relies heavily on visualization and qualitative assessment by expert observers. We propose the use of histogram matching to better assess differences across image formation methods. We briefly demonstrate the technique using a set of sample beamforming methods and discuss the implications of such image processing. We present variations of histogram matching and provide code to encourage the application of this method within the imaging community.
23 citations
Cites background from "Detection performance theory for ul..."
...the CNR, including this resolution term, represents the ideal observer in the task of lesion detection [25], [26]....
[...]
References
More filters
Book•
01 Jan 1965
TL;DR: This chapter discusses the concept of a Random Variable, the meaning of Probability, and the axioms of probability in terms of Markov Chains and Queueing Theory.
Abstract: Part 1 Probability and Random Variables 1 The Meaning of Probability 2 The Axioms of Probability 3 Repeated Trials 4 The Concept of a Random Variable 5 Functions of One Random Variable 6 Two Random Variables 7 Sequences of Random Variables 8 Statistics Part 2 Stochastic Processes 9 General Concepts 10 Random Walk and Other Applications 11 Spectral Representation 12 Spectral Estimation 13 Mean Square Estimation 14 Entropy 15 Markov Chains 16 Markov Processes and Queueing Theory
13,886 citations
Book•
01 Jan 1966
TL;DR: This book discusses statistical decision theory and sensory processes in signal detection theory and psychophysics and describes how these processes affect decision-making.
Abstract: Book on statistical decision theory and sensory processes in signal detection theory and psychophysics
11,820 citations
"Detection performance theory for ul..." refers background or methods in this paper
...The YN ideal observer SNR is related to 2AFC SNR by the equation [5], [24], thus...
[...]
...The test statistic of the ideal observer is the log-likelihood ratio [5], [8]...
[...]
...This type of task is often called a YES/NO (YN) task [5]....
[...]
...Performance is quantified with a receiver operating characteristic (ROC) curve [5]....
[...]
4,028 citations
"Detection performance theory for ul..." refers background in this paper
...These can be obtained from the characteristic function [16] of the test statistic, viewed as the Fourier transform of the probabilities or ....
[...]
TL;DR: Sonography can be used to accurately classify some solid lesions as benign, allowing imaging follow-up rather than biopsy, and this distinction could be definite enough to obviate biopsy.
Abstract: PURPOSE: To determine whether sonography could help accurately distinguish benign solid breast nodules from indeterminate or malignant nodules and whether this distinction could be definite enough to obviate biopsy. MATERIALS AND METHODS: Seven hundred fifty sonographically solid breast nodules were prospectively classified as benign, indeterminate, or malignant. Benign nodules had no malignant characteristics and had either intense homogeneous hyperechogenicity or a thin echogenic pseudocapsule with an ellipsoid shape or fewer than four gentle lobulations. Sonographic classifications were compared with biopsy results. The sensitivity, specificity, and negative and positive predictive values of the classifications were calculated. RESULTS: Benign histologic features were found in 625 (83%) lesions; malignant histologic features, in 125 (17%). Of benign lesions, 424 had been prospectively classified as benign. Two lesions classified as benign were found to be malignant at biopsy. Thus, the classification s...
1,683 citations
"Detection performance theory for ul..." refers background in this paper
...shadowing are important diagnostic features in applications such as breast cancer sonography [12] we focus on echogenicity as shown in Fig....
[...]
TL;DR: Mammographic sensitivity for breast cancer declines significantly with increasing breast density and is independently higher in older women with dense breasts, which significantly increases detection of small cancers and depicts significantly more cancers and at smaller size and lower stage than does PE, which detects independently extremely few cancers.
Abstract: PURPOSE: To (a) determine the performance of screening mammography, ultrasonography (US), and physical examination (PE); (b) analyze the influence of age, hormonal status, and breast density; (c) compare the size and stage of tumors detected with each modality; and (d) determine which modality or combination of modalities optimize cancer detection. MATERIALS AND METHODS: A total of 11,130 asymptomatic women underwent 27,825 screening sessions, (mammography and subsequent PE). Women with dense breasts subsequently underwent screening US. Abnormalities were deemed positive if biopsy findings revealed malignancy and negative if findings from biopsy or all screening examinations were negative. RESULTS: In 221 women, 246 cancers were found. Sensitivity, specificity, negative and positive predictive values, and accuracy of mammography were 77.6%, 98.8%, 99.8%, 35.8%, and 98.6%, respectively; those of PE, 27.6%, 99.4%, 99.4%, 28.9%, and 98.8%, respectively; and those of US, 75.3%, 96.8%, 99.7%, 20.5%, and 96.6%,...
1,591 citations
"Detection performance theory for ul..." refers background in this paper
...In a recent screening study [1], ultrasound in...
[...]