CC Views Research Articles

Utilizing Pseudo Color Image to Improve the Performance of Deep Transfer Learning-Based Computer-Aided Diagnosis Schemes in Breast Mass Classification.

The purpose of this study is to investigate the impact of using morphological information in classifying suspicious breast lesions. The widespread use of deep transfer learning can significantly improve the performance of themammogram based CADx schemes.However, digital mammograms are grayscale images, while deep learning models are typically optimized using the natural images containing three channels. Thus, it is needed to convert the grayscale mammograms into three channel images for the input of deep transfer models. This study aims to developa novel pseudo color image generation methodwhich utilizesthe mass contour information to enhance the classification performance.Accordingly, a total of 830 breast cancer cases were retrospectively collected, which contains 310 benign and 520 malignant cases, respectively. For each case, a total of four regions of interest (ROI) are collected from the grayscale images captured for both the CC and MLO views of the two breasts. Meanwhile, a total of seven pseudo color image sets are generated as the input of the deep learning models, which arecreated through a combination of the original grayscale image, a histogram equalized image, a bilaterally filtered image, and a segmented mass. Accordingly, the output features from four identical pre-trained deep learning models are concatenated and then processed by a support vector machine-based classifier to generate the final benign/malignant labels. The performance of each image set was evaluated and compared. The results demonstrate that the pseudo color sets containing the manually segmented mass performed significantly better than all other pseudo color sets, which achieved an AUC (area under the ROC curve) up to 0.889 ± 0.012 and an overall accuracy up to 0.816 ± 0.020, respectively. At the same time, the performance improvement is also dependent on the accuracy of the mass segmentation. The results of this study support our hypothesis that adding accurately segmented mass contours can provide complementary information, thereby enhancing the performance of the deep transfer model in classifying suspicious breast lesions.

A Survey of Mean Glandular Doses and Suggestions on National Diagnostic Reference Levels for Digital Mammography in China.

The primary purpose of this study was to report the mean glandular doses and to determine the national diagnostic reference levels for digital mammography based on data between 2016 and 2018 in China. The data from 19,076 mammograms (4,769 examinations) by random sampling from 118 digital mammography systems were compiled. Exposure factors included age, compressed breast thickness, kVp, mAs, target/filter combination, entrance surface air kerma, and mean glandular doses, which were retrospectively surveyed and recorded from the monitor. The national diagnostic reference levels (75th percentiles) in mean glandular dose were calculated across median value obtained for all included data and stratified to specific compressed breast thickness ranges. The patients' ages ranged from 22 to 88 y, with a median age of 45. The applied voltage and output medians were 28 kVp and 75.1 mAs for all exposure, respectively. The median CBTs were 45 mm and 48 mm for craniocaudal views and mediolateral oblique views, and the corresponding median mean glandular doses were 1.32 mGy and 1.40 mGy, respectively. The national diagnostic reference level at compressed breast thickness of 40-50 mm was 1.67 mGy for CC views and 1.71 mGy for MLO views. The median mean glandular doses varied significantly and increased with compressed breast thickness, demonstrating the necessity of establishing DRL according to breast thickness and optimizing the clinic's digital mammography practice in China.

Parallel Hierarchical Multi-View Feature Fusion Based on Canonical Correlation Analysis for Mammogram Retrieval

Due to the diversity of image sources, content-based multi-source image fusion and retrieval have shown promising capabilities in computer vision tasks, and especially when applied in Computer-Aided Diagnosis (CAD) to automate and improve the accuracy of medical image analysis. The combination of computer vision and CAD systems has the potential to revolutionize healthcare by augmenting the expertise of clinicians, improving overall diagnostic accuracy and helping experts in the clinical decision-making process by classifying and retrieving similar annotated clinical images to a given query. In the context of multi-view mammography interpretation, the concept of multi-view feature fusion has recently been studied to improve retrieval performance while effectively guaranteeing the complementarity of both MLO and CC views. However, conventional multi-view feature fusion makes descriptors long and lacks to take into consideration the relationship between descriptors. To deal with this issue, we propose two hierarchical multi-view feature fusion methods, for multi-view mammogram retrieval, based on the Canonical Correlation Analysis (CCA), which is the most commonly used multivariate parametric test. In fact, we have adapted CCA to determine the relationship between two descriptors by processing latent correlation factors. Moreover, after extracting descriptors for each view, a comparative study of texture and shape fusion descriptors is proposed in order to identify the more discriminative features for multi-view mammogram retrieval. Then, a query-dependent distance metric preserving both visual resemblance and semantic similarity is carried out to dynamically determine the more appropriate distance measure for each query image. Extensive experiments on the challenging Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM) have demonstrated the effectiveness of the proposed hierarchical multi-view feature fusion for mammogram retrieval, which outperforms the performance achieved either by conventional fused information or by single view information. To improve the transparency of our paper, the source code of the proposed method and the related dataset (including readme files) are publicly accessible through the following GitHub link: https://github.com/ABDERRAHIMMAR/Multi-View-Feature-Fusion-for-Mammogram-Retrieval . This open-access resource empowers researchers and practitioners to delve deeper into our methodology, fostering collaboration and advancements in the field of computer-aided diagnosis and medical image analysis.

A single-institution retrospective evaluation of noninvasive localization for non-palpable breast microcalcification

TechniqueFrom January 1, 2018, to December 31, 2021, we localized the breast microcalcification of 40 patients before the surgical excision. We measured the distance between the nipple and the center of the calcification on the CC view and the ML view, respectively. The operation proceeded around the intersection between two lines, slightly larger than the diameter of the microcalcification. We also analyze the pathological findings. ResultsAll 40 patients successfully detected calcification by mammograms preoperatively using the method mentioned above. 38 patients have the microcalcification removal within the one-time operation, while the other two underwent an extended lumpectomy. 20 of 40 calcifications (50 %) were malignant and 12(30 %) were precancerous lesions. In the group of women older than 45 years old, the percentages of malignant and atypical hyperplasias are 56.25 % (18/32) and 31.25 % (10/32) respectively. ConclusionOur non-invasive method of preoperative localization is safe and cost-effective. Furthermore, initial observations suggest that there may be a link between age and malignant microcalcification.

Saliency of breast lesions in breast cancer detection using artificial intelligence

The analysis of mammograms using artificial intelligence (AI) has shown great potential for assisting breast cancer screening. We use saliency maps to study the role of breast lesions in the decision-making process of AI systems for breast cancer detection in screening mammograms. We retrospectively collected mammograms from 191 women with screen-detected breast cancer and 191 healthy controls matched by age and mammographic system. Two radiologists manually segmented the breast lesions in the mammograms from CC and MLO views. We estimated the detection performance of four deep learning-based AI systems using the area under the ROC curve (AUC) with a 95% confidence interval (CI). We used automatic thresholding on saliency maps from the AI systems to identify the areas of interest on the mammograms. Finally, we measured the overlap between these areas of interest and the segmented breast lesions using Dice’s similarity coefficient (DSC). The detection performance of the AI systems ranged from low to moderate (AUCs from 0.525 to 0.694). The overlap between the areas of interest and the breast lesions was low for all the studied methods (median DSC from 4.2% to 38.0%). The AI system with the highest cancer detection performance (AUC = 0.694, CI 0.662–0.726) showed the lowest overlap (DSC = 4.2%) with breast lesions. The areas of interest found by saliency analysis of the AI systems showed poor overlap with breast lesions. These results suggest that AI systems with the highest performance do not solely rely on localized breast lesions for their decision-making in cancer detection; rather, they incorporate information from large image regions. This work contributes to the understanding of the role of breast lesions in cancer detection using AI.

Diagnostic accuracy of subjective kinetic assessment of masses in contrast-enhanced mammography in comparison with contrast-enhanced magnetic resonance imaging

BackgroundContrast-enhanced MRI (CE MRI) of the breast is currently the most sensitive imaging technique for detecting invasive breast cancer, and it provides both morphologic and functional information through kinetics for characterizing breast masses. Contrast-enhanced mammography (CEM) uses the same principle of neo-angiogenesis to detect early cancers similar to MRI with comparable diagnostic performance. However, there is an important limitation in CEM in characterizing the breast lesions because of the non-availability of kinetic information. To the best of our knowledge, very few studies have assessed the CEM kinetics. In this study, we have evaluated the accuracy of subjective assessment of contrast kinetics in CEM and compared it with the subjective and quantitative kinetic assessment in CE MRI. If the performance of CEM is comparable to MRI, it may add an additional dimension to CEM in characterizing the breast masses in addition to detection.ResultsKinetic information of 123 lesions in 90 patients was analyzed in CEM and MRI. Of these, 26 (21.1%) were benign, 4 (3.3%) were high risk lesions, and 93 (75.6%) were malignant breast lesions. Comparison of subjective and quantitative assessment in CE MRI had almost perfect agreement with a kappa value of 0.816, and both were used as reference standards for comparing CEM kinetics. Comparison of subjective assessment of kinetic patterns in CEM using only CC and MLO views showed moderate agreement with both quantitative (kappa − 0.483) and subjective (0.547) CE MRI kinetics. When the delayed image obtained at 8 min was included for kinetic analysis, CEM kinetics showed substantial to almost perfect agreement with quantitative (kappa − 0.673) and subjective (kappa − 0.855) CE MRI kinetics, respectively.ConclusionWe hope that this study results would encourage the breast radiologist to assess the kinetic information from CEM and use CEM as a single, simple and cost-effective imaging modality in detecting and characterizing breast masses.

Automatic and standardized quality assurance of digital mammography and tomosynthesis with deep convolutional neural networks

ObjectivesThe aim of this study was to develop and validate a commercially available AI platform for the automatic determination of image quality in mammography and tomosynthesis considering a standardized set of features.Materials and methodsIn this retrospective study, 11,733 mammograms and synthetic 2D reconstructions from tomosynthesis of 4200 patients from two institutions were analyzed by assessing the presence of seven features which impact image quality in regard to breast positioning. Deep learning was applied to train five dCNN models on features detecting the presence of anatomical landmarks and three dCNN models for localization features. The validity of models was assessed by the calculation of the mean squared error in a test dataset and was compared to the reading by experienced radiologists.ResultsAccuracies of the dCNN models ranged between 93.0% for the nipple visualization and 98.5% for the depiction of the pectoralis muscle in the CC view. Calculations based on regression models allow for precise measurements of distances and angles of breast positioning on mammograms and synthetic 2D reconstructions from tomosynthesis. All models showed almost perfect agreement compared to human reading with Cohen’s kappa scores above 0.9.ConclusionsAn AI-based quality assessment system using a dCNN allows for precise, consistent and observer-independent rating of digital mammography and synthetic 2D reconstructions from tomosynthesis. Automation and standardization of quality assessment enable real-time feedback to technicians and radiologists that shall reduce a number of inadequate examinations according to PGMI (Perfect, Good, Moderate, Inadequate) criteria, reduce a number of recalls and provide a dependable training platform for inexperienced technicians.

O074 Localisation using multiple magnetic seeds in unilateral breast surgery: a report of clinical and radiological outcomes in a tertiary breast cancer centre

Abstract Introduction Magnetic seeds (Magseeds) are becoming increasingly used as a method of localising non-palpable breast lesions. The cost effectiveness and reduced scheduling conflicts in comparison with wire localization has led to increased popularity in recent years. There's a paucity of data on outcomes when using multiple magseeds in one breast. In this study, we assess the radiological, operative and pathological outcomes of inserting multiple magseeds in the same breast. Methods A retrospective analysis of all magseed guided excisions preformed at the MMUH breast surgery department between January 2020 and December 2022. 40 patients with more than one magseed inserted unilaterally were included. Primary outcome measures were successful excision of lesion(s) and magseed retrieval. Secondary outcomes included re-excision rate and perioperative complications. Results The mean distance between magseeds was 48.3 and 42.5 mm on ML and CC views respectively. 9/40 used magseeds for bracketing while the rest used for excision of discrete lesions. Mean operative time was 56.65 minutes and 55% of the procedures were wide local excisions. There was no reported difficulties identifying magseeds intra-operatively or in the specimen mammogram. The mean specimen size is 33.42 mm while the mean size of the final lesion is 15.49 mm. Invasive ductal carcinoma was the commonest pathological outcome. The rate of positive margins was 20% and re-excision was 15%. One patient proceeded to completion mastectomy. Conclusion Multiple magseed localisation is a feasible, safe and effective method when utilized in bracketing a unifocal lesion or separate lesion excisions in unilateral breast surgery.

Deep-learning convolutional neural network-based scatter correction for contrast enhanced digital breast tomosynthesis in both cranio-caudal and mediolateral-oblique views.

Scatter radiation in contrast-enhanced digital breast tomosynthesis (CEDBT) reduces the image quality and iodinated lesion contrast. Monte Carlo simulation can provide accurate scatter estimation at the cost of computational burden. A model-based convolutional method trades off accuracy for processing speed. The purpose of this study is to develop a fast and robust deep-learning (DL) convolutional neural network (CNN)-based scatter correction method for CEDBT. Projection images and scatter maps of digital anthropomorphic breast phantoms were generated using Monte Carlo simulations. Experiments were conducted to validate the simulated scatter-to-primary ratio (SPR) at different locations of a breast phantom. Simulated projection images were used for CNN training and testing. Two separate U-Nets [low-energy (LE)-CNN and high-energy (HE)-CNN] were trained for LE and HE spectrum, respectively. CNN-based scatter correction was applied to a clinical case with a malignant iodinated mass to evaluate the influence on the lesion detection. The average and standard deviation of mean absolute percentage error of LE-CNN and HE-CNN estimated scatter map are and , respectively. For clinical cases, the lesion signal difference to noise ratio average improvement was 190% after CNN-based scatter correction. To conduct scatter correction on clinical CEDBT images, the whole process of loading CNNs parameters and scatter correction for LE and HE images took , with 9GB GPU computational cost. The SPR variation across the breast agrees between experimental measurements and Monte Carlo simulations. We developed a CNN-based scatter correction method for CEDBT in both CC view and mediolateral-oblique view with high accuracy and fast speed.

A combined deep CNN-lasso regression feature fusion and classification of MLO and CC view mammogram image

A combined deep CNN-lasso regression feature fusion and classification of MLO and CC view mammogram image

Mammographic positioning in women with pectus excavatum: An anatomic challenge

PurposeTo assess mammographic image quality in women with pectus excavatum (PEx) compared to women without PEx. Materials & methodsFifty-six women with PEx between the ages 36–80 (median, 57 years) with screening mammograms from 2006 to 2020 were identified in an IRB-approved HIPAA-compliant retrospective review. Two fellowship-trained breast radiologists independently evaluated mammographic quality of 109 individual breasts in the 56 women using Enhancing Quality Using the Inspection Program (EQUIP) positioning criteria and visual breast density assessments. The number of images per breast was documented. Comparison was made to 2:1 age-matched controls whose screening mammograms were performed in the same year. A power analysis for the difference in the number of images per breast between study groups was performed before data collection. ResultsStatistically significant differences with worse performance in women with PEx included: the pectoralis muscle extending to the posterior nipple line (p < 0.0001); adequacy of tissue visualized (p < 0.0001); inframammary fold included (p < 0.0001); breast free of skin folds (p = 0.003); presence of fibroglandular tissue at the CC view posterior edge (p < 0.0001); and CC and MLO within 1 cm of each other (p < 0.001). The average number of images per breast in the PEx group was greater than the control group (2.94 vs. 2.24, p < 0.0001). ConclusionPEx women more often fail to meet mammographic positioning quality standards and more often require additional views for screening.

Mlo and CC view of feature fusion and mammogram classification using deep convolution neural network

Breast cancer is the most frequent type of cancer in women all over the world. The improvement of computer aided system help the radiologist for the effective analysis and diagnosis of breast cancer. It presents a computational methodology for classifying breast cancer as normal, benign and malignant from CC and MLO views of mammogram image. The proposed strategy consists of feature extraction, multiple view feature fusion and classification. The input images are fed into feature extraction where convolution neural network is applied. The CNN is nicely suitable for both feature extraction, feature fusion and mammogram classification. In this framework, convolution layer, pooling and activation function are used as a feature extraction techniques. After the process of feature extraction, feature fusion is employed by average pooling of CNN. The feature fusion will increase or maximize the relevant information of the breast image. Finally obtained features from the fusion are fed into CNN classifier in which softmax and fully connected layer are employed as a classifier techniques. The proposed work achieves 98.4% of accuracy to classify the breast cancer from MLO and CC views using hybrid feature with CNN classifier.

Transformers Improve Breast Cancer Diagnosis from Unregistered Multi-View Mammograms.

Deep convolutional neural networks (CNNs) have been widely used in various medical imaging tasks. However, due to the intrinsic locality of convolution operations, CNNs generally cannot model long-range dependencies well, which are important for accurately identifying or mapping corresponding breast lesion features computed from unregistered multiple mammograms. This motivated us to leverage the architecture of Multi-view Vision Transformers to capture long-range relationships of multiple mammograms from the same patient in one examination. For this purpose, we employed local transformer blocks to separately learn patch relationships within four mammograms acquired from two-view (CC/MLO) of two-side (right/left) breasts. The outputs from different views and sides were concatenated and fed into global transformer blocks, to jointly learn patch relationships between four images representing two different views of the left and right breasts. To evaluate the proposed model, we retrospectively assembled a dataset involving 949 sets of mammograms, which included 470 malignant cases and 479 normal or benign cases. We trained and evaluated the model using a five-fold cross-validation method. Without any arduous preprocessing steps (e.g., optimal window cropping, chest wall or pectoral muscle removal, two-view image registration, etc.), our four-image (two-view-two-side) transformer-based model achieves case classification performance with an area under ROC curve (AUC = 0.818 ± 0.039), which significantly outperforms AUC = 0.784 ± 0.016 achieved by the state-of-the-art multi-view CNNs (p = 0.009). It also outperforms two one-view-two-side models that achieve AUC of 0.724 ± 0.013 (CC view) and 0.769 ± 0.036 (MLO view), respectively. The study demonstrates the potential of using transformers to develop high-performing computer-aided diagnosis schemes that combine four mammograms.

BREAST CANCER CLASSIFICATION USING DEEP LEARNING MODEL FOR CC AND MLO VIEWS OF MAMMOGRAMS

In this work, we explore a way for analyzing a mammography exam holistically. We present the design and implementation of a deep learning model that takes as input the two mammography views (CC and MLO), as well as all detected breast masses and micro calcifications, and produces an output consisting of a three-class classification of the entire exam: negative (or normal), benign, or malignant findings. The registration of the CC and MLO views of a mammography exam is a difficult task due to the difficulty in estimating the non-rigid deformations that can align these two views, so we propose the deep learning model classification, with the hypothesis that the high-level nature of these features will reduce the need for a low-level matching of the input data.

Evaluation of Breast Galactography Using Digital Breast Tomosynthesis: A Clinical Exploratory Study.

Objectives: To compare the application value of digital breast tomosynthesis (DBT) and full-field digital mammography (FFDM) in breast galactography. Materials and Methods: A total of 128 patients with pathological nipple discharge (PND) were selected to undergo galactography. DBT and FFDM were performed for each patient after injecting the contrast agent; the radiation dose of DBT and FFDM was calculated, and the image quality was evaluated in consensus by two senior breast radiologists. Histopathologic data were found in 49 of the 128 patients. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for both FFDM- and DBT-galactography were calculated using histopathologic results as a reference standard. Data were presented as percentages along with their 95% confidence intervals (CI). Results: The average age of the 128 patients was 46.53 years. The average glandular dose (AGD) of DBT-galactography was slightly higher than that of FFDM-galactography (p < 0.001). DBT-galactography was 30.7% higher than FFDM-galactography in CC view, while DBT-galactography increased by 21.7% compared with FFDM-galactography in ML view. Regarding catheter anatomic distortion, structure detail, and overall image quality groups, DBT scores were higher than FFDM scores, and the differences were significant for all measures (p < 0.05). In 49 patients with pathological nipple discharge, we found that the DBT-galactography had higher sensitivity, specificity, PPV, and NPV (93.3%, 75%, 97.7%, and 50%, respectively) than FFDM-galactography (91.1%, 50%, 95.3%, and 33.3%, respectively). Conclusions: Compared to FFDM-galactography, within the acceptable radiation dose range, DBT-galactography increases the sensitivity and specificity of lesion detection by improving the image quality, providing more confidence for the diagnosis of clinical ductal lesions.

Mammographic Breast Densities among 200 Nigerian Females in Sokoto, North Western Nigeria

Background: Breast density has been reported to be an independent predictor of breast cancer risk. Women with highest mammographic breast density may have four to six fold increased risk of developing cancer. Aim and Objective: This study is aimed at evaluating the various forms of breast densities following screening and diagnostic mammographic examination. Materials and Methods: From December 2010 and November 2012, 200 females had mammographic breast examination. Seventy-seven (77) females had routine screening mammography while 123 had diagnostic mammographic examination in the department with the General Electric (GE) Alpha-RT machine with model number MGF-101(manufactured 2010). All the subjects had to fill a mammographic form consisting of variables like age, sex, occupation, family history of breast cancer, tribe, contraception, parity, and caffeine consumption, history of surgical intervention (lumpectomy, biopsy and/or mastectomy), previous mammography and last child birth. MLO and CC views were done for the breast examination though additional views were occasionally employed. Results: Breast density decreases with increase in age of the female subjects and breast lesions are better detected following mammography on fatty or less dense breast.

Optimization of the Radiation Dose of Digital Breast Tomosynthesis in Opportunistic Screening by Studying the Effect of Different Combinations of FFDM and DBT Views.

BackgroundFull-field digital mammography (FFDM) and digital breast tomosynthesis (DBT) are used separately or in combination to identify small breast lesions. The dose of the examination depends on the density of the breast and the imaging (FFDM or DBT) performed. We have performed a retrospective review of FFDM and DBT in women with denser breasts in order to demonstrate how varying the combination of FFDM and DBT in CC and MLO views affects lesion detection and the average gland dose.MethodsEighty-one patients with dense breast received both full-field digital mammography (FFDM) and DBT bilateral screening. The recorded data included the display rates for small lesions or other positive lesions, the type of breast gland, the average gland dose (AGD), and the compression thickness of different collection methods. ANOVA was used to compare the AGD among different collection combinations, and t-test was used to perform pairwise comparison between groups with the same gland type. The relationship between AGD and compression thickness was analyzed by Pearson linear correlation, and the lesion display rates were compared using Chi-square test.ResultsWe found that AGDs were significantly different among the 6 collection methods (F = 119.06, p<0.01), but were not obviously different between groups with the same gland type (F = 0.848, p>0.05). The types of dense glands were correlated with compression thickness, and the thickness was moderately to strongly correlated with AGD (r=0.596–0.736). The combination of single-view DBT(CC-DBT) and FFDM showed significantly higher mass display rates than the two-view FFDM (p<0.05), while the display rates for other positive lesions were similar (p>0.05).ConclusionOur study showed that in opportunistic screening of patients with small breast masses that can be easily detected by ultrasound, MLO-FFDM+CC-DBT or CC-FFDM+MLO-DBT combinations can better balance the individual average gland dose and detection accuracy. The study result cannot be applied to the detection of non-mass lesions as the numbers are too small.

Diagnostic accuracy of resection margin in specimen radiography: digital breast tomosynthesis versus full-field digital mammography.

We investigated the accuracy of digital breast tomosynthesis compared to full-field digital mammography for evaluating tumor-free resection margins in the intraoperative specimen during breast-conserving surgery, reducing re-excision rates. In total, 170 patients, with proven breast cancer and eligible for breast-conserving surgery, were enrolled. Intraoperative specimens underwent digital mammography and digital breast tomosynthesis. Two breast radiologists, with ten years of experience in breast imaging, in batch mode, evaluated tumor-free resection margins and the distance between the margins and lesion. Histopathological findings were considered the standard of reference. We used the correlation analysis to evaluate the agreement between measures oftumor-free resection margins obtained with digital mammography and the true value (histopathological findings), and between digital breast tomosynthesis and histopathological findings. The size evaluation determined by digital breast tomosynthesis was more accurately correlated with that found by pathology; the calculated Pearson's correlation coefficient of digital breast tomosynthesis and digital mammography to the pathologically determined tumor-free resection margins were 0.92 and 0.79 in CC view and 0.92 and 0.72 in LL view, respectively. Compared with the pathologically determined tumor-free resection margins, the size determined by both imaging modalities was, on average, overestimated. Bland-Altman analysis showed an excellent inter readers agreement. Digital breast tomosynthesis is more accurate in assessment of margin status than digital mammography; it could be a more accurate technique than full-field digital mammography for the intraoperative delineating of tumor resection margins.

Diagnostic mammography in Sokoto: A review of 123 cases

Background: Diagnostic mammography is a radiographic examination of the breast to detect the palpable and nonpalpable lesions. More than 1.15 million women are diagnosed of breast cancer yearly worldwide. Objective: This study is aimed at evaluating the mammographic outcomes among 123 Nigerian females who had the sign and symptoms of breast diseases. Materials and Methods: This is a cross-sectional study (retrospective in nature) of 123 adult females who came for diagnostic mammography between December 2010 and November 2012 at the mammographic suite of radiology department UDUTH. MLO and CC views were done for the breast examination though compression views were occasionally employed. Results: One hundred and twenty-three females had diagnostic mammography, with a minimum age of 30 years. The prevalence of diagnostic mammography within the stated period is 61.5%. The mammographic findings were normal in 62 (50.4%) and abnormal in 61 (49.6%) of the subjects. The abnormal mammographic findings were masses in either or both breasts in 45 participants (36.6%), architectural distortion in either or both breasts in 10 participants (8.1%), isolated calcification in either or both breasts in 4 participants (3.3%), left retracted nipple in 1 (0.8%) participant while another subject (0.8%) had a retracted right nipple. Conclusion: Diagnostic mammography can detect the various forms of breast pathologies which were mostly breast masses, calcifications, and architectural distortions. These findings affirmed the need of routine early screening so that breast diseases can be detected early.

A deep learning model using data augmentation for detection of architectural distortion in whole and patches of images

Breast cancer is now widely known to be the second most lethal disease among women. Computer-aided detection (CAD) systems, deep learning (DL) in particular, have continued to provide significant computational solution in early detection and diagnosis of this disease. Research efforts are advancing novel approaches to improve the performance of DL-based models. Techniques such as data augmentation, varying depth of model, image quality enhancement, and choice of classifier have been proposed to improve performance in the characterization of abnormalities in mammograms. However, no significant progress has been made in applying deep learning techniques to the detection of architectural distortion – a form of abnormalities in breast images. In this research, we propose a novel convolution neural network (CNN) model for the detection of architectural distortion by enhancing its performance using data augmentation technique. We also investigate the performance of the proposed model on different operations of image augmentation. Furthermore, the new model was adapted to detect images presenting the right and left breast presented in MLO and CC views. Similarly, we investigate the performance of our model under the fixed-size region of interests (ROIs) and multi-size whole images inputs. Our method was trained on 5136 ROIs from MIAS, 410 whole images from INbreast, 322 whole images from MIAS, and 55,890 ROIs from DDSM + CBS databases. Performance evaluation of the proposed model in comparison with other state-of-the-art techniques revealed that the model achieved 93.75 % accuracy. This study has, therefore, strengthened the need to leverage data augmentation techniques to enhance the detection of architectural distortion, thereby reducing the rate of advanced cases of breast cancer.