Development and evaluation of a robust algorithm for computer-assisted detection of sentinel lymph node micrometastases.

AIMS: Increasing the sectioning rate for breast sentinel lymph nodes can increase the likelihood of detecting micrometastases. To make serial sectioning feasible, we have developed an algorithm for computer-assisted detection (CAD) with digitized lymph node sections. METHODS AND RESULTS: K-means clustering assigned image pixels to one of four areas in a colourspace (representing tumour, unstained background, counterstained background and microtomy artefacts). Four filters then removed 'false-positive' pixels from the tumour cluster. A set of 43 sections containing tumour (a total of 259 foci) and 59 sections negative for malignancy was defined by two pathologists, using light microscopy, and CAD was applied. For the clinically relevant task of identifying the largest focus in each section (micrometastasis in 22/43 sections), the sensitivity and specificity were 100%. Isolated tumour cells (ITCs) were identified in one slide initially considered to be negative. Identification of all 259 foci yielded sensitivities of 57.5% for ITCs (<0.200 mm), 89.5% for micrometastases, and 100% for larger metastases, with one false-positive. Reduced sensitivity was ascribed to variable staining. Nine additional metastases (<0.01-0.3 mm) that were not initially identified were detected by CAD. CONCLUSIONS: This algorithm is well suited to the task of sentinel lymph node evaluation and may enhance the detection of occult micrometastases.

Increasing specimen coverage using digital whole-mount breast pathology: implementation, clinical feasibility and application in research.

Conventional histopathological evaluation is performed on breast specimens using a highly limited sampling of tissues visualized in a two-dimensional (2D) manner although important tumor measurements are three-dimensional. Here we describe a '3D' technique for whole-mount, whole-specimen processing which reduces conformational change and dramatically increases specimen coverage, based on digitizing whole-specimen, whole-mount (up to 12.7cm×17.8cm) serial sections. We describe hardware and software tools for acquiring, viewing and processing the large image datasets (up to 400GB), validation studies investigating the clinical significance of the additional information gleaned from the 3D approach, and application to radiologic-pathologic correlation and biomarker visualization.

Screen-film mammographic density and breast cancer risk: a comparison of the volumetric standard mammogram form and the interactive threshold measurement methods.

BACKGROUND: Mammographic density is a strong risk factor for breast cancer, usually measured by an area-based threshold method that dichotomizes the breast area on a mammogram into dense and nondense regions. Volumetric methods of breast density measurement, such as the fully automated standard mammogram form (SMF) method that estimates the volume of dense and total breast tissue, may provide a more accurate density measurement and improve risk prediction. METHODS: In 2000-2003, a case-control study was conducted of 367 newly confirmed breast cancer cases and 661 age-matched breast cancer-free controls who underwent screen-film mammography at several centers in Toronto, Canada. Conditional logistic regression was used to estimate odds ratios of breast cancer associated with categories of mammographic density, measured with both the threshold and the SMF (version 2.2beta) methods, adjusting for breast cancer risk factors. RESULTS: Median percent density was higher in cases than in controls for the threshold method (31% versus 27%) but not for the SMF method. Higher correlations were observed between SMF and threshold measurements for breast volume/area (Spearman correlation coefficient = 0.95) than for percent density (0.68) or for absolute density (0.36). After adjustment for breast cancer risk factors, odds ratios of breast cancer in the highest compared with the lowest quintile of percent density were 2.19 (95% confidence interval, 1.28-3.72; P(t) <0.01) for the threshold method and 1.27 (95% confidence interval, 0.79-2.04; Pt = 0.32) for the SMF method. CONCLUSION: Threshold percent density is a stronger predictor of breast cancer risk than the SMF version 2.2beta method in digitized images.

Mammographic density and breast cancer risk: evaluation of a novel method of measuring breast tissue volumes.

BACKGROUND: Mammographic density has been found to be strongly associated with risk of breast cancer. We have assessed a novel method of assessing breast tissue that is fully automated, does not require an observer, and measures the volume, rather than the projected area, of the relevant tissues in digitized screen-film mammogram. METHODS: Sixteen mammography machines in seven locations in Toronto were calibrated to allow the estimation of the proportion of radiologically dense (stromal and epithelial tissue) and nondense (fatty) tissue represented in each pixel of the mammographic image. This information was combined with a measurement of breast thickness to calculate the volumes of these tissues. Women with newly diagnosed breast cancer (cases) identified on these mammography machines during the years 2000 to 2003 were compared with other women of the same age who did not have breast cancer (controls). RESULTS: Three hundred sixty-four cases and 656 controls were recruited, epidemiologic data were collected, screen-film mammograms were digitized and measured using both a computer-assisted thresholding method, and the new measure of the volume of density. After adjustment for other risk factors, the odds ratio for those in the 5th quintile compared with the 1st quintile was 1.98 (95% confidence interval, 1.3-3.1) for the volume measure and 1.86 (95% CI, 1.1-3.0) for the area measurement. After inclusion of the volume and area measures in a predictive model, the volume measure lost significance, whereas the area measure remained significant. CONCLUSIONS: Contrary to our expectations, measurement of the volume of breast tissue did not improve prediction of breast cancer risk.

Accurate estimation of compressed breast thickness in mammography.

The determination of volumetric breast density (VBD) from mammograms requires an accurate knowledge of the thickness of the compressed breast. Previously, the authors described a technique for measuring local thicknesses using optical stereoscopic photogrammetry [A. H. Tyson, G. E. Mawdsley, and M. J. Yaffe, "Measurement of compressed breast thickness by optical stereoscopic photogrammetry," Med. Phys. 36(2), 569-576 (2009)]. Here, the authors describe the use of this tool to guide the development of a simpler, more practical field technique for the estimation of breast thickness and test its accuracy. Phantoms were constructed having similar shapes and compression characteristics to breasts of different sizes. These phantoms were compressed at different forces on several types of mammography units and their thickness under compression was measured using optical stereoscopic photogrammetry at many points of contact with the compression plate. A prediction equation was developed that uses the readout of compressed thickness and compression force provided by the mammography system to estimate local breast thickness. Using this approach, systems can be calibrated to an accuracy of better than 5 mm in thickness using a simple test object compared to an error of up to 15 mm associated with using only the thickness readout of the mammography machine. On the systems tested, the estimated value of VBD obtained using this method is significantly reduced from that determined using the constant thickness reported by the mammography machine.

Optical imaging analysis of microscopic radiation dose gradients in Gafchromic EBT film using a digital microscope.

By providing superior localization and immobilization, stereotactic radiosurgery (SRS) is capable of delivering millimeter spheres of dose to intracranial targets with submillimeter precision. Several authors have proposed new SRS solutions to dramatically reduce beam penumbra to hundreds of microns. These solutions require new quality assurance methods capable of penumbra measurement at the micron scale. This article examines the capability of a digital microscope, with translation stage and associated software, to resolve dose gradients in Gafchromic EBT film at this level. To produce very steep penumbra, films were irradiated in phantom beneath pinhole collimators using lower energy x rays (100 kVp, 300 kVp, and Iridium-192) and minimal geometric penumbra contribution. For film analysis, a method was developed which improved the signal to noise ratio by finding the center of the irradiation spot, generating several radial dose profiles and averaging these to obtain the final off-axis dose profile. Optical density was converted to dose using a calibration curve. The experimentally determined off-axis dose profiles were compared with MCNP Monte Carlo simulations which replicated the irradiation geometry and served to validate our measured data. The measured 80%-20% penumbral widths were 46 microm +/- 26 microm (100 kVp, 2 mm field size), 69 microm +/- m 27 microm (300 kVp, 2 mm field size), and 241 microm +/-31 microm (Ir-192, 1 mm field size). These penumbral widths agreed with Monte Carlo simulations within experimental uncertainty. Our findings suggest that reading Gafchromic EBT films using a digital microscope with translation stage is suitable for the quality assurance of very sharp penumbra able to resolve gradients to within at least 30 microm.

Spatial resolution requirements for acquisition of the virtual screening slide for digital whole-specimen breast histopathology.

We examined the effect of lateral spatial resolution and reader specialty on the accuracy of detection of breast cancer. The motivation for this pilot study was the need to acquire and display very large data sets in whole-specimen 3D digital breast histopathology imaging. The ultimate goal is to determine the minimum resolution adequate for detection of malignancy. Twenty-three histologic slides were selected from breast pathology cases and digitized at 2 sampling distances (3.2 and 1.9 microm pixels). Images were viewed by 14 pathologists, of whom 5 had breast pathology as their primary specialty. The readers assessed the likelihood of malignancy on a 5-point Likert scale, and provided a provisional diagnosis. For the detection task, sensitivity, specificity, overall accuracy of detection, and area under the receiver-operator curve were calculated. An overall diagnostic score, and scores grouped by malignancy type, were also computed. Outcome measures were examined for significant resolution and specialty effects. Increasing the lateral resolution significantly improved accuracy in diagnosis (P=.004) but no effect was found for detection. Breast specialists achieved significantly higher scores for all outcome measures except specificity. Differences in performance between the 2 groups of readers tended to be greater for the diagnostic task compared to detection, especially at the higher resolution. However, specimen coverage may also be a significant factor. Factors related to the readers may have also affected performance in this study. Based on these results, a more comprehensive study should examine pixel sizes between 0.7 and 1.9 microm.

Design and characterization of a digital image acquisition system for whole-specimen breast histopathology.

We have developed a digital histopathology imaging system capable of producing a three-dimensional (3D) representation of histopathology from an entire lumpectomy specimen. The system has the potential to improve the accuracy of surgical margin assessment in the treatment of breast cancer by providing finer sampling and 3D visualization. A scanning light microscope was modified to allow digital photomicrography of a stack of large (up to 120x170 mm2) histology slides cut serially through the entire specimen. The images are registered and displayed in 2D and 3D. The design of the system, which reduces or eliminates the appearance of 'tiling' and 'seam' artefacts inherent in the scanning method, is described and its resolution, contrast/noise and coverage properties are characterized through measurements of the modulation transfer function (MTF), depth of field (DOF) and signal difference to noise ratio (SDNR). The imaging task requires a lateral resolution of 5 microm, an SDNR of 5 between relevant features, 'tiling artefact' at a level below the detectability threshold of the eye, and 'seam artefact' of less than 5-10 microm. The tests demonstrate that the system is largely adequate for the imaging task, although further optimizations are required to reduce the degradation of coverage incurred by seam artefact.