KEYWORDS: Data modeling, Interference (communication), Visual process modeling, Visualization, Electronic filtering, Receivers, Statistical analysis, Medical imaging, Data acquisition, Performance modeling
The binormal receiver operating characteristic (ROC) model often predicts an unphysical "hook" near the upperright
corner (1,1) of the ROC plot. Several models for fitting proper ROC curves avoid this problem. The purpose of
this work is to describe another method that involves a model of visual search that models free-response data, and to
compare the search-model predicted ROC curves with those predicted by PROPROC (proper ROC) software. The
highest rating rule was used to infer ROC data from FROC data. An expression for the search-model ROC
likelihood function is derived, maximizing which yielded estimates of the parameters and the fitted ROC curve. The
method was applied to a dual-modality 5-reader FROC data set. The relative difference between the average AUCs
for the two methods was less than 1%. A linear regression of the AUCs yielded an adjusted R-squared of 0.95
indicative of strong linear correlation between the search model AUC and PROPROC AUC, although the shapes of
the predicted ROC curves were qualitatively different. This study shows the feasibility of estimating parameters
characterizing visual search from data acquired in a non-search paradigm.
The purpose of this study was to investigate the effect of dose on lesion detection and characterization in breast
tomosynthesis (BT), using human breast specimens. Images of 27 lesions in breast specimens were acquired on a BT
prototype based on a Mammomat Novation (Siemens) full-field digital mammography (FFDM) system. Two detector
modes - binned (2×1 in the scan direction) and full resolution - and four BT exposure levels - approximately 2×, 1.5×,
1×, and 0.5× the total mAs at the same beam quality as used in a single FFDM view with a Mammomat Novation unit
under automatic exposure control (AEC) conditions - were examined. The exposure for all BT scans was equally
divided among 25 projections. An enhanced filtered back projection reconstruction method was applied with a constant
filter setting. A human observer performance study was conducted in which the observers were forced to select the
minimum (threshold) exposure level at which each lesion could be both detected and characterized for assessment of
recall or not in a screening situation. The median threshold exposure level for all observers and all lesions corresponded
to approximately 1×, which is half the exposure of what we currently use for BT. A substantial variation in exposure
thresholds was noticed for different lesion types. For low contrast lesions with diffuse borders, an exposure threshold of
approximately 2× was required, whereas for spiculated high contrast lesions and lesions with well defined borders, the
exposure threshold was lower than 0.5×. The use of binned mode had no statistically significant impact on observer
performance compared to full resolution mode. There was no substantial difference between the modes for the detection
and characterization of the lesion types.
The purpose of this work was to develop a contrast-detail phantom that can be used to evaluate image quality in breast
tomosynthesis (BT) and as a first step use it to evaluate in-plane artifacts with respect to object size and contrast. The
phantom was constructed using a Polylite® resin as bulk material, as it has x-ray mass attenuation properties similar to
polymethyl methacrylate (PMMA), a common phantom material in mammography. Six different materials -
polyoxymethylene (POM), bakelite®, nylon, polycarbonate (PC), acrylonitrilebutadienestyrene (ABS) and polyethene
(PE) - were selected to form the phantom details. For each of the six materials, five spherical objects were manufactured
(diameters of 4, 8, 12, 16, and 20 mm) resulting in 30 objects that were embedded with their centres approximately
aligned at the central plane of a 26 mm thick Polylite® block (210 mm x 300 mm). A 20 mm thick PMMA block was
added to yield a phantom with attenuation properties similar to 45 mm PMMA that could simulate a so-called standard
breast (50 mm thick, 50% glandular tissue). Images of the phantom were acquired using a BT prototype system that
employs filtered backprojection for image reconstruction. The magnitude of the in-plane artifacts was evaluated and was
found to increase linearly with increasing contrast (signal) level and size of the embedded objects. The contrast-detail
phantom was found to be a useful tool for evaluating BT in-plane artifacts and might also be used to study out-of-plane
artifacts and the effect of different acquisition and reconstruction parameters on image quality in BT.
The purpose of this study was to determine how image quality in breast tomosynthesis (BT) is affected when acquisition
modes are varied, using human breast specimens containing malignant tumors and/or microcalcifications. Images of
thirty-one breast lumpectomy and mastectomy specimens were acquired on a BT prototype based on a Mammomat
Novation (Siemens) full-field digital mammography system. BT image acquisitions of the same specimens were
performed varying the number of projections, angular range, and detector signal collection mode (binned and nonbinned
in the scan direction). An enhanced filtered back projection reconstruction method was applied with constant
settings of spectral and slice thickness filters. The quality of these images was evaluated via relative visual grading
analysis (VGA) human observer performance experiments using image quality criteria. Results from the relative VGA
study indicate that image quality increases with number of projections and angular range. A binned detector collecting
mode results in less noise, but reduced resolution of structures. Human breast specimens seem to be suitable for
comparing image sets in BT with image quality criteria.
The purpose of this work was to evaluate and compare the visibility of tumors in digital mammography (DM) and breast tomosynthesis (BT) images. Images of the same women were acquired on both a DM system (Mammomat Novation, Siemens) and a BT prototype system adapted from the same type of DM system. Simulated 3D tumors (average dimension: 8.4 mm x 6.6 mm x 5 mm) were projected and added to each DM image as well as each BT projection image prior to 3D reconstruction. The same beam quality and approximately the same total absorbed dose were used for each breast image acquisition on both systems. Two simulated tumors were added to each of thirty breast scans, yielding sixty cases. A series of 4-alternative forced choice (4-AFC) human observer performance experiments were conducted in order to determine what projected tumor signal intensity in the DM images would be needed to achieve the same detectability as in the reconstructed BT images. Nine observers participated. For the BT experiment, when the tumor signal intensity on the central projection was 0.010 the mean percent of correct responses (PC) was measured to be 81.5%, which converted to a detectability index value (d') of 1.96. For the DM experiments, the same detectability was achieved at a signal intensity determined to be 0.038. Equivalent tumor detection in BT images were thus achieved at around four times less projected signal intensity than in DM images, indicating that the use of BT may lead to earlier detection of breast cancer.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.