Microcalcifications In Digital Mammography Research Articles

Visual and quantitative evaluation of microcalcifications in mammograms with deep learning-based super-resolution

PurposeTo evaluate visually and quantitatively the performance of a deep-learning-based super-resolution (SR) model for microcalcifications in digital mammography. MethodMammograms were consecutively collected from 5080 patients who underwent breast cancer screening from January 2015 to March 2017. Of these, 93 patients (136 breasts, mean age, 50 ± 7 years) had microcalcifications in their breasts on mammograms. We applied an artificial intelligence model known as a fast SR convolutional neural network to the mammograms. SR and original mammograms were visually evaluated by four breast radiologists using a 5-point scale (1: original mammograms are strongly preferred, 5: SR mammograms are strongly preferred) for the detection, diagnostic quality, contrast, sharpness, and noise of microcalcifications. Mammograms were quantitatively evaluated using a perception-based image-quality evaluator (PIQE). ResultsAll radiologists rated the SR mammograms better than the original ones in terms of detection, diagnostic quality, contrast, and sharpness of microcalcifications. These ratings were significantly different according to the Wilcoxon signed-rank test (p <.001), while the noise score of the three radiologists was significantly lower (p <.001). According to PIQE, SR mammograms were rated better than the original mammograms, showing a significant difference by paired t-test (p <.001). ConclusionAn SR model based on deep learning can improve the visibility of microcalcifications in mammography and help detect and diagnose them in mammograms.

Microcalcification diagnosis in digital mammography using extreme learning machine based on hidden Markov tree model of dual-tree complex wavelet transform

Abstract Diagnosis of benign and malignant microcalcifications in digital mammography using Computer-aided Diagnosis (CAD) system is critical for the early diagnosis of breast cancer. Wavelet transform based diagnosis methods are effective to accomplish this task, but limited by representing the correlation within each wavelet scale, these methods neglect the correlation between wavelet scales. In this paper, we apply the hidden Markov tree model of dual-tree complex wavelet transform (DTCWT-HMT) for microcalcification diagnosis in digital mammography. DTCWT-HMT can effectively capture the correlation between different wavelet coefficients and model the statistical dependencies and non-Gaussian statistics of real signals, is used to characterize microcalcifications for the diagnosis of benign and malignant cases. The combined features which consist of the DTCWT-HMT features and the DTCWT features are optimized by genetic algorithm (GA). Extreme learning machine (ELM), an efficient learning theory is employed as the classifier to diagnose the benign and malignant microcalcifications. The validity of the proposed method is evaluated on the Nijmegen, MIAS and DDSM datasets using area under curve (AUC) of receiver operating characteristic (ROC). The AUC values of 0.9856, 0.9941 and 0.9168 of the proposed method are achieved on Nijmegen, MIAS and DDSM, respectively. We compare the proposed method with state-of-the-art diagnosis methods, and the experimental results show the effectiveness of the proposed method for the diagnosis of the benign and malignant microcalcifications in mammograms in terms of the accuracy and stability.

A theoretical investigation of spectra utilization for a CMOS based indirect detector for dual energy applications

Dual Energy imaging is a promising method for visualizing masses and microcalcifications in digital mammography. Currently commercially available detectors may be suitable for dual energy mammographic applications. The scope of this work was to theoretically examine the performance of the Radeye CMOS digital indirect detector under three low- and high-energy spectral pairs. The detector was modeled through the linear system theory. The pixel size was equal to 22.5μm and the phosphor material of the detector was a 33.9 mg/cm2 Gd2O2S:Tb phosphor screen. The examined spectral pairs were (i) a 40kV W/Ag (0.01cm) and a 70kV W/Cu (0.1cm) target/filter combinations, (ii) a 40kV W/Cd (0.013cm) and a 70kV W/Cu (0.1cm) target/filter combinations and (iii) a 40kV W/Pd (0.008cm) and a 70kV W/Cu (0.1cm) target/filter combinations. For each combination the Detective Quantum Efficiency (DQE), showing the signal to noise ratio transfer, the detector optical gain (DOG), showing the sensitivity of the detector and the coefficient of variation (CV) of the detector output signal were calculated. The second combination exhibited slightly higher DOG (326 photons per X-ray) and lower CV (0.755%) values. In terms of electron output from the RadEye CMOS, the first two combinations demonstrated comparable DQE values; however the second combination provided an increase of 6.5% in the electron output.

Modeling indirect detectors for performance optimization of a digital mammographic detector for dual energy applications

Dual Energy imaging is a promising method for visualizing masses and microcalcifications in digital mammography. The advent of two X-ray energies (low and high) requires a suitable detector. The scope of this work is to determine optimum detector parameters for dual energy applications. The detector was modeled through the linear cascaded (LCS) theory. It was assumed that a phosphor material was coupled to a CMOS photodetector (indirect detection). The pixel size was 22.5 μm. The phosphor thickness was allowed to vary between 20mg/cm2 and 160mg/cm2 The phosphor materials examined where Gd2O2S:Tb and Gd2O2S:Eu. Two Tungsten (W) anode X-ray spectra at 35 kV (filtered with 100 μm Palladium (Pd)) and 70 kV (filtered with 800 pm Ytterbium (Yb)), corresponding to low and high energy respectively, were considered to be incident on the detector. For each combination the contrast- to-noise ratio (CNR) and the detector optical gain (DOG), showing the sensitivity of the detector, were calculated. The 40 mg/cm2 and 70 mg/cm2 Gd2O2S:Tb exhibited the higher DOG values for the low and high energy correspondingly. Higher CNR between microcalcification and mammary gland exhibited the 70mg/cm2 and the 100mg/cm2 Gd2O2S:Tb for the low and the high energy correspondingly.

Using breast radiographers' reports as a second opinion for radiologists' readings of microcalcifications in digital mammography.

The aim of this study was to investigate a practical method for incorporating radiographers' reports with radiologists' readings of digital mammograms. This simulation study was conducted using data from a free-response receiver operating characteristic observer study obtained with 75 cases (25 malignant, 25 benign and 25 normal cases) of digital mammograms. Each of the rating scores obtained by six breast radiographers was utilized as a second opinion for four radiologists' readings with the radiographers' reports. A logical "OR" operation with various criteria settings was simulated for deciding an appropriate method to select a radiographer's report in all combinations of radiologists and radiographers. The average figure of merit (FOM) of the radiologists' performances was statistically analysed using a jackknife procedure (JAFROC) to verify the clinical utility of using radiographers' reports. Potential improvement of the average FOM of the radiologists' performances for identifying malignant microcalcifications could be expected when using radiographers' reports as a second opinion. When the threshold value of 2.6 in Breast Imaging-Reporting and Data System (BI-RADS®) assessment was applied to adopt/reject a radiographer's report, FOMs of radiologists' performances were further improved. When using breast radiographers' reports as a second opinion, radiologists' performances potentially improved when reading digital mammograms. It could be anticipated that radiologists' performances were improved further by setting a threshold value on the BI-RADS assessment provided by the radiographers. For the effective use of a radiographer's report as a second opinion, radiographers' rating scores and its criteria setting for adoption/rejection would be necessary.

Impact of Magnification Views on the Characterization of Microcalcifications in Digital Mammography

To evaluate the additional benefit of true geometric (air-gap) magnification views for the characterization of microcalcifications in digital mammography. After ethical approval, we retrospectively reviewed patient records to identify 100 patients with suspicious microcalcifications (35 malignant, 65 benign) who had a standard digital mammography and an additional digital magnification view in the same projection within three months. All images were obtained using an amorphous silicon-based full-field digital system (Senographe 2000 D, GE Healthcare, Chalfont St. Giles, UK). Images were independently analyzed by six board-certified radiologists. The probability of malignancy was estimated using first standard contact mammography alone (MG) and then mammography in combination with the magnification view (MG+MAG) using a modified Breast Imaging Reporting and Data System (BI-RADS) classification system and a percentage scale. Results were compared using receiver operating characteristic (ROC) analysis. In addition, readers assessed the subjective visibility of the calcifications. For all six readers combined, the area under the curve (AUC) was 0.664 ± 0.052 for MG and 0.813 ± 0.042 for MG + MAG, resulting in a statistically significant improvement of 0.148 ± 0.120. Each reader had a higher AUC for MG + MAG than MG, with the improvement being statistically significant in four of the six readers. In 76.34 % of the cases, MG + MAG resulted in better visibility of calcifications compared with mammography alone. In 33 % slightly more and in 39 % significantly more calcifications were found. Even in digital mammography with the option of using electronic magnification (zoom) at the viewing workstation, true geometric (air-gap) magnification views remain important for the visibility and correct classification of microcalcifications and for the assessment of their extent.

SU‐FF‐I‐114: Effects of Exposure Level and Anatomical Background On Detection of Simulated Microcalcifications in Digital Mammography

Purpose: To assess the influence of quantum and anatomic noises on the detection of simulated microcalcifications (MCs) on digital mammograms. Materials and Methods: Images of an anthropomorphic breast phantom (RMI 169) overlapping with simulated MCs ranging from 160 to 224 micron in size placed at various locations were acquired with an aSi/aSe flat panel based digital mammography system (Selenia) operated with Mo‐Mo target/filter combination at 28 kVp. These mammograms were exposed with 10, 20, 40, 80, 160, and 325 mAs. All images were displayed randomly on a review workstation and reviewed by five readers independently. The readers were asked to count the number of visible MCs. The ratio of visible MCs was computed for each different combination of MC size, reader and mAs setting. The performance differences in the ratios averaged from all readers between mAs levels was assessed by performing Student t‐test to compute the p values. Results: It was found that in the dense tissue region the visibility reached 90% at 10, 33 and 37 mAs for 200–212, 160–180 and 180–200 micron MCs, respectively; and at 10 mAs for 180–200 and 200–212 micron MCs and at 15 mAs for 160–180 micron MCs in the fatty region. The lower visibility in the dense tissue region may be due to the fact that the photon flux was lower in the dense tissue regions therefore requiring higher mAs setting to achieve sufficient image SNRs for detection of MCs. Our results also show that in the dense tissue region, the visibility of 180–200 micron MCs stayed at 90% at 80 mAs. This may indicate that the anatomical noise was limiting the detectability of 180–200 micron MCs.This work was supported in part by grants CA104759 and CA124585 from NIH‐NCI, a grant EB00117 from NIH‐NIBIB, and a subcontract from NIST‐ATP.

A quantitative method for evaluating the detectability of lesions in digital mammography

This study presents a quantitative method for evaluating the detectability of microcalcifications in digital mammography. Four hundred and twenty microcalcifications (with various morphology, size and contrast), simulated with a previously validated method, were used for the creation of image datasets. Lesions were inserted into 163 regions of interests of 59 selected raw digital mammograms with various anatomical backgrounds and acquired with a Siemens Novation DR. After processing, these composite images were scored by experienced radiologists, who located multiple simulated lesions and rated them under conditions of free-search. For statistical analysis, free-response receiver-operating characteristic curves are plotted; the use of jackknife free-response receiver-operating characteristic method has also been investigated. The main advantage of this methodology is that the exact number of inserted microcalcifications is well known and that the lesions are fully characterised in terms of pathology, size, morphology and peak contrast. A first application has been the evaluation of the effect of anatomical background on microcalcifications detection. Preliminary findings in this study indicate that this method may be a promising tool to evaluate factors that have an influence on the detectability of lesions, such as the clinical processing or the viewing conditions.

Threshold pixel size for shape determination of microcalcifications in digital mammography: a pilot study

The effect of pixel size on shape determination in screening digital mammography systems was studied using a shape identification task as the measured outcome. Ten microcalcifications on screen-films were digitised to a range of pixel sizes (2.5-200 microm) and extracted from computed radiography (CR) images (50 microm) acquired under equivalent imaging conditions. Fifteen observers attempted to identify the shape of each microcalcification at each pixel size. The results were collated to provide a fraction of correct responses vs. pixel size curve for each microcalcification. Averaging over all shapes, pixel values >100 microm lead to a significant decrease in shape determination ability (p < 0.01) for digitised screen-film. For CR images, half the shapes were not properly identified. Hence, although 20-100 microm was sufficient for microcalcification shape determination for digitised screen-film images, 50 microm was only borderline sufficient for the CR digital images.

A Spatial Scan Statistic for Stochastic Scan Partitions

This article develops a spatial scan statistic for homogeneity analysis of point processes that utilizes stochastic scan partitions. The derivation of the sampling distribution for the statistic yields an exact test. This test has the potential for improved power over conventional alternatives when the point process is embedded in an underlying continuous random field and is recommended in situations for which the location of subregions of nonhomogeneity in the point process correspond to regions in the underlying field that can be segmented as distinct from their surroundings. The application to the detection of clustered microcalcifications in digital mammography is investigated as a motivating example.

Tree structured wavelet transform segmentation of microcalcifications in digital mammography.

A novel multistage algorithm is proposed for the automatic segmentation of microcalcification clusters (MCCs) in digital mammography. First, a previously reported tree structured nonlinear filter is proposed for suppressing image noise, while preserving image details, to potentially reduce the false positive (FP) detection rate for MCCs. Second, a tree structured wavelet transform (TSWT) is applied to the images for microcalcification segmentation. The TSWT employs quadrature mirror filters as basic subunits for both multiresolution decomposition and reconstruction processes, where selective reconstruction of subimages is used to segment MCCs. Third, automatic linear scaling is then used to display the image of the segmented MCCs on a computer monitor for interpretation. The proposed algorithms were applied to an image database of 100 single view mammograms at a resolution of 105 microns and 12 bits deep (4096 gray levels). The database contained 50 cases of biopsy proven malignant MCCs, 8 benign cases, and 42 normal cases. The measured sensitivity (true positive detection rate) was 94% with a low FP detection rate of 1.6 MCCs/image. The image details of the segmented MCCs were reasonably well preserved, for microcalcification of less than 500 microns, with good delineation of the extent of the microcalcification clusters for each case based on visual criteria.

Digital Mammography

We investigated the spatial resolution requirement and the effect of unsharp-mask filtering on the detectability of subtle microcalcifications in digital mammography. Digital images were obtained by digitizing conventional screen-film mammograms with a 0.1 X 0.1 mm2 pixel size, processed with unsharp masking, and then reconstituted on film with a Fuji image processing/simulation system (Fuji Photo Film Co., Tokyo, Japan). Twenty normal cases and 12 cases with subtle microcalcifications were included. Observer performance experiments were conducted to assess the detectability of subtle microcalcifications in the conventional, the unprocessed digital, and the unsharp-masked mammograms. The observer response data were evaluated using receiver operating characteristic (ROC) and LROC (ROC with localization) analyses. Our results indicate that digital mammograms obtained with 0.1 X 0.1 mm2 pixels provide lower detectability than the conventional screen-film mammograms. The detectability of microcalcifications in the digital mammograms is improved by unsharp-mask filtering; the processed mammograms still provide lower accuracy than the conventional mammograms, however, chiefly because of increased false-positive detection rates for the processed images at each subjective confidence level. Viewing unprocessed digital and unsharp-masked images in pairs resulted in approximately the same detectability as that obtained with the unsharp-masked images alone. However, this result may be influenced by the fact that the same limited viewing time was necessarily divided between the two images.