Mammographic Detection Research Articles

Artificial Intelligence (AI)-Based Computer-Assisted Detection and Diagnosis for Mammography: An Evidence-Based Review of Food and Drug Administration (FDA)-Cleared Tools for Screening Digital Breast Tomosynthesis (DBT)

Artificial Intelligence (AI)-Based Computer-Assisted Detection and Diagnosis for Mammography: An Evidence-Based Review of Food and Drug Administration (FDA)-Cleared Tools for Screening Digital Breast Tomosynthesis (DBT)

The Effectiveness of Semi-Supervised Learning Techniques in Identifying Calcifications in X-ray Mammography and the Impact of Different Classification Probabilities

Identifying calcifications in mammograms is crucial for early breast cancer detection, and semi-supervised learning, which utilizes a small dataset for supervised learning combined with deep learning, is anticipated to be an effective approach for automating this identification process. This study explored the impact of semi-supervised learning on identifying mammographic calcifications by including 712 mammographic images from 252 patients in public datasets. Initially, 212 mammogram images were segmented into patches and classified visually for calcification presence. A subset of these patches, derived from 169 mammogram images, was used to train a ResNet50-based classifier. The classifier was evaluated using patches generated from 43 mammograms as a test data set. Additionally, 500 more mammogram images were processed into patches and analyzed using the trained ResNet50 model, with semi-supervised learning applied to patches exceeding certain classification probabilities. This process aimed to enhance the classifier’s accuracy and achieve improvements over the initial model. The findings indicated that semi-supervised learning significantly benefits the accuracy of calcification detection in mammography, underscoring its utility in enhancing diagnostic methodologies.

USPSTF Recommendations and Overdiagnosis.

Overdiagnosis is the concept that some cancers detected at screening would never have become clinically apparent during a woman's lifetime in the absence of screening. This could occur if a woman dies of a cause other than breast cancer in the interval between mammographic detection and clinical detection (obligate overdiagnosis) or if a mammographically detected breast cancer fails to progress to clinical presentation. Overdiagnosis cannot be measured directly. Indirect methods of estimating overdiagnosis include use of data from randomized controlled trials (RCTs) designed to evaluate breast cancer mortality, population-based screening studies, or modeling. In each case, estimates of overdiagnosis must consider lead time, breast cancer incidence trends in the absence of screening, and accurate and predictable rates of tumor progression. Failure to do so has led to widely varying estimates of overdiagnosis. The U.S. Preventive Services Task Force (USPSTF) considers overdiagnosis a major harm of mammography screening. Their 2024 report estimated overdiagnosis using summary evaluations of 3 RCTs that did not provide screening to their control groups at the end of the screening period, along with Cancer Intervention and Surveillance Network modeling. However, there are major flaws in their evidence sources and modeling estimates, limiting the USPSTF assessment. The most plausible estimates remain those based on observational studies that suggest overdiagnosis in breast cancer screening is 10% or less and can be attributed primarily to obligate overdiagnosis and nonprogressive ductal carcinoma in situ.

Abstract PO2-13-12: Serum Thioredoxin 1 Enhances Breast Cancer Detection in Mammography and Breast Ultrasound, Including Dense Breasts

Abstract Background: Breast cancer (BC), when detected in its early stages, generally carries a more favorable prognosis than other types of cancer. Currently, the screening and early detection of breast cancer rely predominantly on image-oriented diagnostic methods. However, these methods inherently fall short in capturing the dynamic nature of cancer cells. Mammography has known limitations in clinical sensitivity, particularly in cases with high breast density. While breast ultrasound can mitigate some of these limitations, there remain unmet needs in this field. Previously, we reported the promising clinical performance (Sensitivity: 96.43%, Specificity: 97.32%) of serum thioredoxin 1 (Trx1) as a potential novel means for detecting breast cancer. In this study, we aim to evaluate the extent to which serum Trx1 can enhance the accuracy of breast cancer diagnosis, including in dense breasts, within the imaging-based diagnostic system powered by mammography and ultrasonography. Methods: We have generated monoclonal antibodies against Trx1 and developed an ELISA kit (DxMe® BC Kit, E&S Healthcare, Korea) that quantifies Trx1 in serum. A total of 308 biopsy-confirmed breast cancer patients, who had undergone mammography prior to their final diagnosis, were recruited. The clinical information of these 308 patients was retrospectively examined and collected. The data from 254 individuals, who also had results from breast ultrasound, and 261 individuals, who had records of breast density, were analyzed collectively. Freshly collected blood was prepared as previously indicated, and then tested with the kit to determine the level of Trx1. All serum samples were measured twice, and statistical analyses were conducted using ROC analysis, one-way ANOVA, and unpaired t-tests. Results: Clinical sensitivity for BC detection was analyzed using the Trx1 levels of the 254 subjects who underwent both mammography and breast ultrasound. The respective sensitivities of mammography, breast ultrasound, and the DxMe® BC Kit were 72.83%, 92.91%, and 96.46%. With combined analysis of mammography, breast ultrasound, and Trx1, the sensitivity increased to 98.43% for mammography + breast ultrasound, and 98.82% for mammography + Trx1. When breast ultrasound + Trx1 was considered, the sensitivity rose to 99.61%. Likewise, the sensitivity of mammography + breast ultrasound + Trx1 was also measured at 99.61%. Among the 261 breast cancer patients with available information about breast density, 186 patients (71.3%) had dense breasts. The clinical sensitivity of mammography and Trx1 was specifically analyzed for these dense breast cases. Mammography demonstrated a sensitivity of 64.3%, while the DxMe® BC kit exhibited a sensitivity of 97.2%. When mammography and Trx1 were analyzed in parallel, sensitivity increased to 99.7%. Conclusion: This study demonstrated that the DxMe® BC kit, which quantifies serum Trx1 levels, can enhance the accuracy of conventional image-based diagnostics for breast cancer screening and early detection by more than 25%. This improvement in sensitivity was observed when mammography and breast ultrasound were each used in conjunction with Trx1, with no significant difference noted between these two combinations. Notably, in the case of dense breasts, the Trx1 test improved sensitivity by more than 30% compared to mammography. Given that sensitivity increases when mammography and Trx1 tests are used together, it showed that incorporating the Trx1 test alongside traditional mammography could significantly enhance diagnostic performance in dense breasts up to 35%. This approach has the potential to streamline the breast cancer diagnostic process, reduce time and cost, and improve the reliability of current breast cancer screening and early detection methods, regardless of breast density. Citation Format: Young Kim, Hyemi Ko, Yeri Lee, Keunkoo Shin, Jongam Song, Jeryoung Kim, Kyounghoon Suh, Jinsun Lee. Serum Thioredoxin 1 Enhances Breast Cancer Detection in Mammography and Breast Ultrasound, Including Dense Breasts [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-13-12.

Prediction and analysis of breast cancer based on transfer learning from DDSM mammography images

Breast cancer is a cancer mostly found in women, causing a high morality rate. It appears as a lump in the breast initially, and as cancer cells spread, the other organs are endangered. Research shows that the earlier the breast cancer is detected, the higher rate of recovering will be. The main methods of breast cancer detection are mainly using various physical methods to probe the internal condition of the patients breast. Breast X-ray imaging, also known as mammography, is the most widely used. It uses low-dose X-ray technology to observe the breast tissue and help doctors figure out the exact condition of the patient. With the advancements in deep learning and image recognition, the analysis of mammograms can be previously done by the computer, easing the pressure on doctors. Digital Database for Screening Mammography (DDSM) dataset is a mammography dataset specifically designed for the development of mammography detection models. It contains huge amounts of mammograms and provides the labels corresponding to them. In this study, the performances of normal deep learning model and transfer learning model on the DDSM dataset are compared. Pretrained neural network models are employed and fine-tuned on the DDSM dataset to adapt to the features of mammographic images. Experimental results demonstrate significant improvements through transfer learning, outperforming traditional methods with Convolutional Neural Network (CNN). Finally, the models are evaluated in multiple aspects, conclusions and prospects are analysed based on the results.

The effect of polycystic ovarian syndrome on fibrocystic breast changes in postmenopausal women

Background Fibrocystic breast changes (FCCs) are benign lesions thought to be caused by an increased estrogen-to-progesterone ratio. One of the most common endocrinopathies that increases this ratio is polycystic ovarian syndrome (PCOS). Although nonproliferative FCCs do not increase the risk of breast cancer, they can make mammographic detection of malignancy in postmenopausal women more difficult. The aim of this study was to investigate the effects of PCOS on the development of postmenopausal FCCs. Methods This retrospective cohort study used the TriNetX research network to identify two cohorts of postmenopausal women (Z78.0) older than 45, without a prior diagnosis of FCCs (N60.1) or hormone replacement therapy (Z79.890). One cohort included a diagnosis of PCOS (E28.2). The cohorts were balanced for age, race, ethnicity, and hormonally relevant comorbidities. The cohorts were then evaluated for the development of FCCs after menopause. Results Postmenopausal patients with PCOS were 52% more likely to develop FCCs than those without PCOS (2.2% vs. 1.4%, relative risk 1.52, 95% confidence interval 1.05, 2.22, P = 0.03). Conclusion Postmenopausal women with PCOS have a higher risk of developing FCCs. Further studies are needed to improve the differentiation of benign FCCs from malignant lesions on imaging for postmenopausal women with PCOS who develop FCCs.

Comparative analysis of automated breast volume scanner (ABVS) combined with conventional hand-held ultrasound and mammography in female breast cancer detection.

This study aimed to compare the diagnostic value of an automated breast volume scanner (ABVS) combined with conventional hand-held ultrasound and mammography in detecting female breast cancer. Early detection is vital in improving patient outcomes for this prevalent disease. Seventy-eight suspicious breast lesions from 60 patients were examined between August 2019 and July 2020. Each patient underwent ABVS, conventional hand-held ultrasound, and mammography. Diagnostic values, including coincidence rate, sensitivity, specificity, positive predictive value, and negative predictive value, were calculated using histopathology results as the "gold standard." Histopathology confirmed 55 malignant (70.51%) and 23 benign lesions (29.48%). ABVS combined with conventional hand-held ultrasound identified 56 malignant (52 confirmed, 4 benign) and 22 benign nodules (3 confirmed, 19 benign). Mammography detected 48 malignant (45 confirmed, 3 benign) and 30 benign nodules (10 confirmed, 20 benign). ABVS combined with conventional hand-held ultrasound had a sensitivity of 94.5%, specificity of 82.6%, positive predictive value of 92.9%, and negative predictive value of 86.4%. Mammography showed a sensitivity of 81.8%, specificity of 87.0%, positive predictive value of 93.8%, and negative predictive value of 66.7%. ABVS combined with conventional hand-held ultrasound showed high diagnostic value in detecting female breast cancer. The "convergence sign" in the coronal section played a significant role. It slightly outperformed mammography and offered advantages in terms of cost, convenience, comfort, and absence of radiation. Further promotion and implementation are supported.

Comprehensive diagnosis of advanced-stage breast cancer: exploring detection methods, molecular subtypes, and demographic influences – A cross-sectional study

BackgroundBrazil faces notable Breast Cancer (BC) mortality despite lower incidence rates versus developed countries. Despite guidelines from medical societies, Brazilian public policy recommends biennial mammographic screening for women aged 50 to 69. This study investigates sociodemographic and clinical factors related to BC detection methods and clinical stage at diagnosis. MethodsThe authors conducted a cross-sectional study at a São Paulo tertiary hospital. Patients were divided into ‘symptomatic’ and ‘mammographic’ detection groups. Bivariate analyses by detection method and clinical stage compared groups' profiles in terms of sociodemographic and clinical characteristics. Poisson regression analyses assessed sociodemographic and molecular subtypes´ influence on “mammographic detection” prevalence and “advanced-stage BC”, reporting prevalence ratios and 95 % Confidence Intervals. ResultsThe authors studied 1,536 BC patients admitted from January 2016 to December 2017. The “mammographic detection” group had a higher proportion of patients aged 50‒69 years (62.9 % vs. 44.1 %), white race (63.3 % vs. 51.6 %), Catholic religion (58.2 % vs. 51.1 %), and Luminal A subtype (25.2 % vs. 13.2 %) compared to the “symptomatic detection” group. Patients with early-stage disease were more likely to have higher education levels (8.1 % vs. 5.5 %) and be married (39.8 % vs. 46.6 %) compared to those with advanced-stage. Molecular subtypes were significantly associated with the detection method and stage. The prevalence of advanced-stage disease in "mammographic" (n=313) and "symptomatic" (n=1191) groups was 18.5 % and 55 %, respectively . Mammographic detection significantly reduced advanced-stage BC prevalence (PR = 0.40, 95 % CI 0.31‒0.51). ConclusionMammographic detection reduces advanced-stage breast cancer prevalence in Brazil, emphasizing the importance of regular screenings, especially among at-risk sociodemographic groups. Enhancing mammographic screening accessibility, lowering the starting age to 40, and extending coverage to include annual mammograms can significantly lower breast cancer mortality in Brazil, benefiting public health and patient outcomes.

Assessment of the Iraqi Breast Cancer Early Detection and Downstaging Program: Mammography Cancer Detection Rate

Background: In Iraq, early detection and downstaging of breast cancer programs were launched to control breast cancer (BC) in 2000. In the program model, mammography is conducted for normal and symptomatic women. The aim of this study is to assess the program’s model effectiveness in terms of mammography cancer detection rates and other screening determinates in the Iraqi Early Detection Program. Method: This is a retrospective study based on data attained from the Breast Disease Treatment Centre (BDTC) in Sulaimani, Iraq. From September 2016 to August 2021, 35,045 women visited the centre for BC screening and diagnosis of breast disorders. In BDTC, 16,186 mammograms and 1,289 biopsies were conducted, and 404 women were diagnosed with BC. This study measured the percentage of women who visited BDTC for screening, the women’s ages at diagnosis, the cancer detection rate (BC per 1,000 women visiting), mammography detection rate (BC per 1,000 mammograms) and the percentage of positive core biopsies. Results: The cancer detection rate (CDR) in 1,000 women was 11.53, the overall mammography detection rate was 24.96, and the highest rate was observed in 2018 (42.2). The overall percentage of positive core biopsies was 31.34%. The highest percentage biopsy rate was in 2017 (43.84%). From 2016 to 2021, the percentage of visits for screening increased gradually, rising from 9.5% to 28.6%. Conclusion: The early detection program in Iraq is considered to be effective in terms of high mammography cancer detection rates and the percentage of positive biopsy. Compared to the previous data, more women participated in the program for screening than for diagnosis.

Application of Artificial Intelligence in the Mammographic Detection of Breast Cancer in Saudi Arabian Women

Background: Breast cancer has a 14.8% incidence rate and an 8.5% fatality rate in Saudi Arabia. Mammography is useful for the early detection of breast cancer. Researchers have been developing artificial intelligence (AI) algorithms for early breast cancer diagnosis and reducing false-positive mammography results. The aim of this study was to examine the performance and accuracy of an AI system in breast cancer screening among Saudi women. Materials and Methods: This is a retrospective cross-sectional study that included 378 mammograms collected from 2017 to 2021 from government hospitals in Jeddah, Saudi Arabia. The patients’ demographic and clinical information were collected from files and electronic medical records. The radiologists’ assessments of the mammograms were based on Breast Imaging Reporting and Data System (BIRADS) scores. Follow-up or biopsy reports verified the radiologists’ findings. The MammoScreen system was the AI tool used in this study. Data were analyzed using SPSS Version 25. Results: The patients’ mean age was 50.31 years. Most patients had breast density B (42.3%) followed by A (27.2%) and C (25.9%). Most malignant cases were invasive ductal carcinomas (37.3%). Of the 181 cancer cases, 36.9% were BIRADS category V. The area under the curve for the AI detection (0.923; 95% confidence interval [CI], 0.893–0.954) was greater than that for the radiologists’ interpretation (0.838; 95% CI, 0.796–0.881). The AI detection agreed with the histopathological result in 167 positive (91.3%) and 182 negative cases (93.3%). The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of the AI system were 92.8%, 91.9%, 91.3%, 93.3%, and 92.3%, respectively. The radiologist’s interpretation agreed with the pathology report in 180 positive (73.8%) and 134 negative cases (100%). Its sensitivity, specificity, PPV, NPV, and accuracy were 100%, 67.7%, 73.8%, 100%, and 83.1%, respectively. Conclusions: The AI system tested in this study had better accuracy and diagnostic performance than the radiologists and thus could be used as a support diagnostic tool for breast cancer detection in clinical practice and to reduce false-positive recalls.

Improving Mass Detection in Mammography Using Focal Loss Based RetinaNet

Breast cancer is a significant global health issue and plays a crucial role in improving patient outcomes through early detection. This study aims to enhance the accuracy and efficiency of breast cancer diagnosis by investigating the application of the RetinaNet and Faster R-CNN algorithms for mass detection in mammography images. A specialized dataset was created for mass detection from mammography images and validated by an expert radiologist. The dataset was trained using RetinaNet and Faster R-CNN, a state-of-the-art object detection model. The training and testing were conducted using the Detectron2 platform. To avoid overfitting during training, data augmentation techniques available in the Detectron2 platform were used. The model was tested using the AP50, precision, recall, and F1-Score metrics. The results of the study demonstrate the success of RetinaNet in mass detection. According to the obtained results, an AP50 value of 0.568 was achieved. The precision and recall performance metrics are 0.735 and 0.60 respectively. The F1-Score metric, which indicates the balance between precision and recall, obtained a value of 0.66. These results demonstrate that RetinaNet can be a potential tool for breast cancer screening and has the potential to provide accuracy and efficiency in breast cancer diagnosis. The trained RetinaNet model was integrated into existing PACS (Picture Archiving and Communication System) systems and made ready for use in healthcare centers.

Reproducibility and Explainability of Deep Learning in Mammography: A Systematic Review of Literature.

Background Although abundant literature is currently available on the use of deep learning for breast cancer detection in mammography, the quality of such literature is widely variable. Purpose To evaluate published literature on breast cancer detection in mammography for reproducibility and to ascertain best practices for model design. Methods The PubMed and Scopus databases were searched to identify records that described the use of deep learning to detect lesions or classify images into cancer or noncancer. A modification of Quality Assessment of Diagnostic Accuracy Studies (mQUADAS-2) tool was developed for this review and was applied to the included studies. Results of reported studies (area under curve [AUC] of receiver operator curve [ROC] curve, sensitivity, specificity) were recorded. Results A total of 12,123 records were screened, of which 107 fit the inclusion criteria. Training and test datasets, key idea behind model architecture, and results were recorded for these studies. Based on mQUADAS-2 assessment, 103 studies had high risk of bias due to nonrepresentative patient selection. Four studies were of adequate quality, of which three trained their own model, and one used a commercial network. Ensemble models were used in two of these. Common strategies used for model training included patch classifiers, image classification networks (ResNet in 67%), and object detection networks (RetinaNet in 67%). The highest reported AUC was 0.927 ± 0.008 on a screening dataset, while it reached 0.945 (0.919-0.968) on an enriched subset. Higher values of AUC (0.955) and specificity (98.5%) were reached when combined radiologist and Artificial Intelligence readings were used than either of them alone. None of the studies provided explainability beyond localization accuracy. None of the studies have studied interaction between AI and radiologist in a real world setting. Conclusion While deep learning holds much promise in mammography interpretation, evaluation in a reproducible clinical setting and explainable networks are the need of the hour.

Automated breast cancer detection in mammography using ensemble classifier and feature weighting algorithms

Automated breast cancer detection in mammography using ensemble classifier and feature weighting algorithms

Artificial intelligence for breast cancer detection in screening mammography in Sweden: a prospective, population-based, paired-reader, non-inferiority study

Artificial intelligence (AI) as an independent reader of screening mammograms has shown promise, but there are few prospective studies. Our aim was to conduct a prospective clinical trial to examine how AI affects cancer detection and false positive findings in a real-world setting. ScreenTrustCAD was a prospective, population-based, paired-reader, non-inferiority study done at the Capio Sankt Göran Hospital in Stockholm, Sweden. Consecutive women without breast implants aged 40-74 years participating in population-based screening in the geographical uptake area of the study hospital were included. The primary outcome was screen-detected breast cancer within 3 months of mammography, and the primary analysis was to assess non-inferiority (non-inferiority margin of 0·15 relative reduction in breast cancer diagnoses) of double reading by one radiologist plus AI compared with standard-of-care double reading by two radiologists. We also assessed single reading by AI alone and triple reading by two radiologists plus AI compared with standard-of-care double reading by two radiologists. This study is registered with ClinicalTrials.gov, NCT04778670. From April 1, 2021, to June 9, 2022, 58 344 women aged 40-74 years underwent regular mammography screening, of whom 55 581 were included in the study. 269 (0·5%) women were diagnosed with screen-detected breast cancer based on an initial positive read: double reading by one radiologist plus AI was non-inferior for cancer detection compared with double reading by two radiologists (261 [0·5%] vs 250 [0·4%] detected cases; relative proportion 1·04 [95% CI 1·00-1·09]). Single reading by AI (246 [0·4%] vs 250 [0·4%] detected cases; relative proportion 0·98 [0·93-1·04]) and triple reading by two radiologists plus AI (269 [0·5%] vs 250 [0·4%] detected cases; relative proportion 1·08 [1·04-1·11]) were also non-inferior to double reading by two radiologists. Replacing one radiologist with AI for independent reading of screening mammograms resulted in a 4% higher non-inferior cancer detection rate compared with radiologist double reading. Our study suggests that AI in the study setting has potential for controlled implementation, which would include risk management and real-world follow-up of performance. Swedish Research Council, Swedish Cancer Society, Region Stockholm, and Lunit.

Use of novel artificial intelligence computer-assisted detection (AI-CAD) for screening mammography: an analysis of 17,884 consecutive two-view full-field digital mammography screening exams.

Novel artificial intelligence computer-assisted detection (AI-CAD) systems based on deep learning (DL) promise to support screen reading. To test a DL-AI-CAD system compared to human reading on consecutive screening mammograms. In this retrospective study, 17,884 consecutive anonymized screening mammograms, double-read from January to November 2018, were processed by the DL-AI-CAD system. AI-CAD reading was considered positive if the AI-CAD case scores exceeded 30 (range = 1-100) and the lesion was correctly marked. Likewise, human reading (R1 or R2, respectively) was considered positive if the lesion was correctly identified and called. Receiver operating characteristic (ROC) analysis was performed and accuracy data were calculated. Ground truth for benign lesions: absence of malignancy after cancer registry matching (2022); for malignancy: histopathologic proof; evaluation was patient-based. In total, 114 screen-detected and 17 interval cancers (ICA) occurred. ROC analysis of screen-detected cancers yielded an AUC of 89% for AI-CAD. Sensitivity/specificity was 81.7%/80.2% for AI-CAD; 77.1%/91.7% for R1; 78.6/91.6% for R2. Combining each human reading with AI-CAD was as sensitive as human double-reading (all approximately 88%), but less specific (approximately 75%) compared to human double-reading (approximately 87%). These AI-CAD combinations required consensus readings for twice as many cases as the human combination. Four of 17 ICA exceeded a case score of 30; two of four CAD correctly marked the quadrant of the subsequent ICA. Including ICA cases, this AI-CAD achieved comparable sensitivity to human reading at lower specificity. Combining human reading and AI-CAD allows increasing sensitivity compared to single-reading.

Enhanced breast mass mammography classification approach based on pre-processing and hybridization of transfer learning models.

The second most prevalent cause of death among women is now breast cancer, surpassing heart disease. Mammography images must accurately identify breast masses to diagnose early breast cancer, which can significantly increase the patient's survival percentage. Although, due to the diversity of breast masses and the complexity of their microenvironment, it is still a significant issue. Hence, an issue that researchers need to continue searching into is how to establish a reliable breast mass detection approach in an effective factor application to increase patient survival. Even though several machine and deep learning-based approaches were proposed to address these issues, pre-processing strategies and network architectures were insufficient for breast mass detection in mammogram scans, which directly influences the accuracy of the proposed models. Aiming to resolve these issues, we propose a two-stage classification method for breast mass mammography scans. First, we introduce a pre-processing stage divided into three sub-strategies, which include several filters for Region Of Interest (ROI) extraction, noise removal, and image enhancements. Secondly, we propose a classification stage based on transfer learning techniques for feature extraction, and global pooling for classification instead of standard machine learning algorithms or fully connected layers. However, instead of using the traditional fine-tuning feature extraction phase, we proposed a hybrid model where we concatenate two recent pre-trained CNNs to assist the feature extraction phase, rather than using one. Using the CBIS-DDSM dataset, we managed to increase mainly each of the accuracy, sensitivity, and specificity reaching the highest accuracy of 98,1% using the Median filter for noise removal. Followed by the Gaussian filter trial with 96% accuracy, meanwhile, the winner filter attained the lowest accuracy of 94.13%. Moreover, the usage of global average pooling as a classifier is suitable in our case better than global max pooling. The experimental findings demonstrate that the suggested strategy of breast Mass detection in mammography can outperform the top-ranked methods currently in use in terms of classification performance.

Artificial Intelligence Applications in Breast Imaging: Current Status and Future Directions.

Attempts to use computers to aid in the detection of breast malignancies date back more than 20 years. Despite significant interest and investment, this has historically led to minimal or no significant improvement in performance and outcomes with traditional computer-aided detection. However, recent advances in artificial intelligence and machine learning are now starting to deliver on the promise of improved performance. There are at present more than 20 FDA-approved AI applications for breast imaging, but adoption and utilization are widely variable and low overall. Breast imaging is unique and has aspects that create both opportunities and challenges for AI development and implementation. Breast cancer screening programs worldwide rely on screening mammography to reduce the morbidity and mortality of breast cancer, and many of the most exciting research projects and available AI applications focus on cancer detection for mammography. There are, however, multiple additional potential applications for AI in breast imaging, including decision support, risk assessment, breast density quantitation, workflow and triage, quality evaluation, response to neoadjuvant chemotherapy assessment, and image enhancement. In this review the current status, availability, and future directions of investigation of these applications are discussed, as well as the opportunities and barriers to more widespread utilization.

Abstract 5561: Molecular imaging of calcification and soft tissue defines prognosis of DCIS to invasive breast cancer

Abstract Introduction: Ductal carcinoma in situ (DCIS) is a non-obligatory precursor to invasive breast cancer (IBC). Currently there are no reliable markers to indicate if DCIS will progress IBC after an initial treatment, or if it will remain as DCIS. Because of this uncertainty, standard treatment for DCIS includes surgery with or without radiotherapy. Breast microcalcifications are a common feature in the mammographic detection of DCIS and occur in around 80% of DCIS and 50% of IBC. Despite this, their potential role in the development of DCIS and with progression to IBC remains largely unexplored. The main aim of the study is to identify novel biomarkers for DCIS prognosis, based on chemical and molecular compositional changes of calcifications and their surrounding soft tissue in DCIS. Methods: Consecutive tissue section from 422 patients were analyzed using mid-infrared and Raman hyperspectral imaging and X-ray diffraction (XRD) and included (i) ‘DCIS controls’ (DCIS without recurrence) (n=193), (ii) ‘DCIS cases’ (DCIS which subsequently recurred as an invasive disease) (n=123), (iii) ‘synchronous DCIS with invasive cancer’ (n=44) and ‘benign’ (n=62) groups. Both parametrized and data mining approaches including cluster analysis, unsupervised and supervised multivariate statistical analysis and biomodelling based on leave-one-sample-out cross-validation were used to classify the data and develop DCIS prediction models. Results: Prediction modelling of 314 calcification and soft tissue spectral images from 170 patients (‘DCIS controls’=118 versus ‘DCIS cases’=52) showed an AUROC mean value of 0.85 in distinguishing DCIS that did not recur from DCIS that recurred as IBC. Prediction modelling of XRD data from 124 patients (‘DCIS controls’=67 versus ‘DCIS cases’=57) showed an AUROC mean value of 0.80. The calcification spectral features showed pathology specific changes in phosphate to carbonate ratio as well as changes in magnesium whitlockite content. The epithelial features showed changes in protein secondary structure and content indicating structural remodeling between the two groups. XRD parametrized data analysis corroborated these observations in terms of difference in whitlockite, as well as crystallinity/unit axis of the measured calcifications. The prediction models are currently being tested independently on the remaining sample data. Conclusion: Complementary chemical and molecular analysis of breast calcifications and soft tissue using spectroscopy and XRD imaging predicts the likely progression of DCIS to IBC. Subject to the ongoing validation of biomarkers, these approaches appear as promising DCIS risk assessment tools that can inform diagnosis, prognosis and treatment options. Citation Format: Jayakrupakar Nallala, Doriana Calabrese, Sarah Gosling, Esther Lips, Rachel Factor, Allison Hall, Sarah Pinder, Ihssane Bouybayoune, Elinor Sawyer, Lorraine King, Jeffrey Marks, Thomas Lynch, Donna Pinto, Jelle Wesseling, Shelley Hwang, Keith Rogers, Nick Stone, Grand Challenge PRECISION consortium. Molecular imaging of calcification and soft tissue defines prognosis of DCIS to invasive breast cancer. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5561.

Chaotic fractals: Why chaos is the dynamic of carcinogenesis

We can see the fractals in our environment every day (trees, snowflakes, broccoli, etc.). Even the shapes of the DNA helix and anatomical structures are fractal, for example, the branching of blood vessels, bronchi, and cell membranes [1]. Unlike euclidean geometry, fractal geometry reveals how an object with irregularities in many dimensions can be identified by examining how the number of features in one dimension relates to the number of similarly shaped features in other dimensions [2]. Mandelbrot used fractal geometry to describe such irregular shapes and demonstrated that this geometry was an appropriate mathematical language for describing chaotic systems [1]. In fractal geometry, the fractal dimension (FD) is a statistical quantity that gives an indication of how completely a fractal appears to fill space, as one zooms down to finer and finer scales. The FD provides a measure of the complexity of a structure. Increased FD is an indicator of chaos [3]. A complex adaptive system (CAS) is a type of system that is composed of many interacting components, called agents, which can adapt and change their behavior based on their interactions with the environment and with other agents. CAS are characterized by their ability to self-organize and evolve over time, often resulting in emergent properties and behaviors that cannot be predicted from the properties of the individual agents alone. Examples of CAS include ecosystems, economies, social networks, and the human brain. It is also worth noting that a CAS can have both chaotic and regular behavior depending on the circumstances and the complexity of the system. Stem cells can also be considered CAS because they possess many of the characteristics that define CAS. Stem cells have the ability to self-renew, differentiate into multiple cell types, and respond to signals from their environment [4]. Some studies suggest an important role of the feedback loop between cancer cells and the microenvironment. Also, putting cells into an “inappropriate” microenvironmental context can otherwise trigger pathological issues, and even neoplastic transformation [5]. Cancer has previously been demonstrated to be a chaotic behavior of the stem cell [6]. The FD of chromatin has been demonstrated to increase during carcinogenesis and tumor growth in diffuse large B-cell lymphoma, chronic lymphocytic leukemia, oropharyngeal carcinoma, and hepatocarcinoma compared to equivalent normal tissue. A research study of over 3000 cancer specimens revealed the prevalence of fractal chromatin structure in neoplasias, as well as the importance of this arrangement in the creation of chromosomal abnormalities [7]. Fractal analysis of the cell surface is a rather sensitive method that has been recently introduced to characterize cell progression toward cancer. Analysis of FD of cell surface imaged with atomic force microscopy (AFM) showed strong segregation between normal and malignant human cervical epithelial cells [8]. There is a growing literature that shows FD to be a useful measure of the pathologies of the tumoral vascular architecture, and tumor/parenchymal border [3, 9]. Entropy-based fractal image modeling has been used to increase the diagnostic accuracy of mammographic detection of breast tumors. Mammographic density correlates with the FD of breast cancer and eventually with a higher growth rate. Breast masses with higher FD show increased aggressiveness and poor prognosis [5]. The FD of the lung tumor on contrast-enhanced computed tomography (CT) images demonstrated that the use of FD as a predictive indicator of therapeutic response or progression is warranted [10]. Similarly, the FD of liver mass in contrast-enhanced CT is a useful prognostic biomarker for hepatocellular carcinoma patients treated with sunitinib [11]. The FD analysis of thyroid ultrasound images is used for the prediction and early detection of thyroid malignancy [12]. Moreover, fractal geometry is applied in many radiographic analyses. At least, the application of chaos theory opens up the opportunity for novel dynamic access to carcinogenesis. Of course, the application of chaos theory cannot solve all problems, but such an interdisciplinary approach may increase the understanding of carcinogenesis. Mesut Tez: Writing—review and editing (equal). None. The author declares no conflict of interest. Not applicable. Not applicable. If no data are available, there is no data for this review.

Abstract P1-05-21: Change in breast cancer detection method, stage at diagnosis and treatment during the COVID-19 Pandemic: 2019-2021

Abstract Objective: Identify changes in breast cancer detection method, stage at diagnosis and treatment prior to, during and after stay-at-home orders and restricted health care access due to COVID-19. Methods: Statistical comparison of detection method (patient (PtD), mammography (MamD) or other), Anatomic TNM Stage 8 (0-IV) and invasive BC treatment change over time by three time periods (time 1: 2019+Q1 2020; time 2: Q2-Q4 2020; time 3: 2021) using chi-square analysis in an institutional retrospective cohort of first primary breast cancer (BC) patients (n=1799), years 2019-2021. Results: In the years prior to the study, 2016-2019, there was no difference in detection method or stage at diagnosis by year with 682 to 733 newly diagnosed BC annually (p=.462). In 2020 (n=535) and 2021 (n=582) annual diagnosed cases dropped 22% and 15% from 2019 levels. Compared to time 1, time 2 MamD BC dropped significantly (64% to 58%) with a subsequent increase in MamD BC to 70% in time 3 (p < .001) creating a U-shaped curve for MamD over time. PtD BC increased in time 2 from 30% to 36% but declined in time 3 to 25% (p <.001). Concurrently, stage at diagnosis shifted from time 1 to time 2 with stage 0 and I declining [stage 0: 21% to 16%, stage I: 40% to 38%] and stage II and IV increasing [stage II: 28% to 33%, stage IV: 2% to 4% stage IV] (p<.001). Subsequently in 2021 stage shifted again with an increase in stage 0 to 22% and stage I to 45% and a decline in stage II (33% to 24%), III (9% to 7%) and IV (4% to 2%) (p<.001). Combining stage 0 and I, the percentage of lower stage BC declined from 61% to 54% and increased to 67% in time 3 (2021) when health services became more readily accessible. There was no change in type of surgery for invasive breast cancer (stage I-III, n=1386) with equivalent numbers of breast conserving surgery (58%), subcutaneous mastectomy (24%) and mastectomy (18%) over the time period. Chemotherapy treatment rates for invasive BC did not change (38%). Radiation therapy increased from 66% (time 1: 2019+Q1 2020) to 73% (time 2: Q2-Q4 2020) then back to 64% in time 3: 2021 (p=.007) independent of surgery type but concordant with an increase in stage IA and stage IIB BC among invasive breast cancer cases in time 2: Q2-Q4 2020 (p<.001). Likewise, neoadjuvant therapy increased and then declined from 33% to 38% to 29% from time 1 to time 3 (p<.001). Conclusions: Number of diagnosed BC cases fell after the first quarter of 2020 during the time of COVID-19 related shut downs and decreased access to health services. During the Q2-Q4 2020 time period mammography detected BC declined with a relative increase in patient detected breast cancer. When mammography detection declined, BC stage at diagnosis shifted to higher stage concurrent with increased rates of radiation and neoadjuvant therapy. In 2021, the relative increase in mammography detected BC indicates a return to more normal screening patterns with a catch up for screening lost in the prior year due to access limitations. In the third time period: 2021, with the return to prior levels of mammography detected breast cancer, stage shifted back to pre-pandemic expected distribution and the excess treatment with radiation and neoadjuvant therapy declined to previously observed levels. Although the changes in detection method, stage and treatment did not persist they were statistically significant and could represent a need for re-establishing pre-pandemic screening behavior. Table 1 Detection Method by COVID Pandemic Diagnosis Time Periods: 2019-2021 (N=1799). Citation Format: Judith A. Malmgren, Mary Atwood, Henry Kaplan. Change in breast cancer detection method, stage at diagnosis and treatment during the COVID-19 Pandemic: 2019-2021 [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P1-05-21.