A Deep Analysis of Transfer Learning Based Breast Cancer Detection Using Histopathology Images

Automated detection and grading of Invasive Ductal Carcinoma breast cancer using ensemble of deep learning models

Background Earlier detection of breast cancer through mammography screening has reduced disease-specific mortality; however, confounding issues such as technical challenges, breast density, and tumor size can result in false negatives and ultimately later stage diagnosis. Next generation liquid biopsy has the potential to complement mammography and enable earlier detection for more women. We have previously demonstrated high sensitivity and specificity for early detection of invasive breast cancer (IBC) by utilizing a novel category of cancer-associated small RNAs, termed orphan noncoding RNAs (oncRNAs), through a liquid biopsy platform. Here, we further improve the ability to detect breast cancer in a larger, multi-source cohort through an AI-driven approach and demonstrate potential for detection of ductal carcinoma in-situ (DCIS). Methods We utilized The Cancer Genome Atlas (TCGA) small RNA-seq database to discover a library of 20,538 oncRNAs, through a female-specific analysis, that were significantly enriched among 1,103 breast tumors compared to 349 normal tissue samples spanning multiple tissue sites. The diagnostic performance of these oncRNAs were assessed in an independent cohort of archived serum samples from 663 female individuals, sourced from Indivumed (Hamburg, Germany), Proteogenex (Inglewood, CA), and MT Group (Los Angeles, CA), including 279 breast cancer patients of various stages (221 IBC and 58 DCIS; mean age: 57.0 ± 13.8 years; ever-smoker: 25.8%) and 304 age-matched controls (mean age: 58.5 ± 13.9 years; ever-smoker: 23.4%) without breast cancer. All samples were collected between 2010–2022 at time of diagnosis for breast cancer patients. We sequenced the small RNA content of these samples at an average depth of 25.28 ± 9.37 million 50-bp single-end reads. We detected 18,025 (87.8%) unique breast cancer-specific oncRNA species within at least one sample from the study cohort. We then trained a generative AI model using 5-fold cross-validation to predict cancer status for all samples. Results Our oncRNA-based model achieved an overall AUC of 0.95 (95% CI, 0.93–0.97) for prediction of IBC versus cancer-free controls with a sensitivity of 0.87 (0.82–0.91) at 90% specificity. We observed high sensitivities, also at 90% specificity, across all tumor stages and tumor sizes (Table 1). Sensitivities for the earliest stage and smallest tumor size were 0.87 (0.78–0.93) and 0.81 (0.61–0.93) for Stage I (n=83) and T1a–b ( >1mm to ≤10mm; n=26), respectively. Additionally, in a small single-source cohort, we also saw high model accuracy and sensitivity for DCIS, which we aim to confirm in additional cohorts. While our overall cancer cohort primarily consisted of individuals with luminal breast cancer, our model had high sensitivities across all breast cancer subtypes at 0.90 (0.84–0.94), 0.73 (0.59–0.85), and 0.86 (0.42–1.0) for luminal (n=181), HER2 positive (n=49), and triple negative (n=7), respectively. Conclusions We further demonstrate the potential utility of oncRNAs as a blood-based biomarker using an AI algorithm for sensitive and accurate early detection of breast cancer in a large cohort. Additionally, we have shown that this oncRNA-based assay performs well in detecting small, early-stage invasive breast tumors, with potential to detect precursors of breast cancer. Table 1: Model sensitivity in breast cancer by tumor stage and size For each tumor stage and size, as defined by the AJCC 7th Edition breast cancer staging system, sensitivity and 95% Pearson-Clopper confidence intervals (CI) are reported at 90% specificity for the number of samples (N). Citation Format: Noura Tbeileh, Taylor Cavazos, Mehran Karimzadeh, Jeffrey Wang, Alice Huang, Dung Ngoc Lam, Seda Kilinc, Jieyang Wang, Xuan Zhao, Andy Pohl, Helen Li, Lisa Fish, Kimberly Chau, Marra Francis, Lee Schwartzberg, Patrick Arensdorf, Hani Goodarzi, Fereydoun Hormozdiari, Babak Alipanahi. Cell-free orphan noncoding RNAs and AI enable early detection of invasive breast cancer and ductal carcinoma in-situ [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 PO2-13-08.

Utility of Breast Magnetic Resonance Imaging in Determining Candidacy for Partial Breast Irradiation

Breast cancer is a common cause of female mortality in developing countries. Early detection and treatment are crucial for successful outcomes. Breast cancer develops from breast cells and is considered a leading cause of death in women. This disease is classified into two subtypes: invasive ductal carcinoma (IDC) and ductal carcinoma in situ (DCIS). The advancements in artificial intelligence (AI) and machine learning (ML) techniques have made it possible to develop more accurate and reliable models for diagnosing and treating this disease. From the literature, it is evident that the incorporation of MRI and convolutional neural networks (CNNs) is helpful in breast cancer detection and prevention. In addition, the detection strategies have shown promise in identifying cancerous cells. The CNN Improvements for Breast Cancer Classification (CNNI-BCC) model helps doctors spot breast cancer using a trained deep learning neural network system to categorize breast cancer subtypes. However, they require significant computing power for imaging methods and preprocessing. Therefore, in this research, we proposed an efficient deep learning model that is capable of recognizing breast cancer in computerized mammograms of varying densities. Our research relied on three distinct modules for feature selection: the removal of low-variance features, univariate feature selection, and recursive feature elimination. The craniocaudally and medial-lateral views of mammograms are incorporated. We tested it with a large dataset of 3002 merged pictures gathered from 1501 individuals who had digital mammography performed between February 2007 and May 2015. In this paper, we applied six different categorization models for the diagnosis of breast cancer, including the random forest (RF), decision tree (DT), k-nearest neighbors (KNN), logistic regression (LR), support vector classifier (SVC), and linear support vector classifier (linear SVC). The simulation results prove that our proposed model is highly efficient, as it requires less computational power and is highly accurate.

Introduction: Breast cancer is the leading cause of cancer deaths in the Brazilian female population, and it is the most common malignant tumor in women in the world. The only method that has proven to decrease breast cancer mortality is mammography and, therefore, screening programs are based on this test in several countries. However, its sensitivity in the general population ranges from 75% to 80%, being especially lower in the case of dense breasts, more common in young women, ranging from 30% to 48% in this group. The most sensitive method in the detection of breast cancer is magnetic resonance imaging, (80% to 97.8%, according to current studies), due to its ability of studying vascular changes in the tissues. Aiming to combine the morphological study provided by mammography with the analysis of tumor perfusion allowed by studies that use intravenous contrasts, contrast enhanced digital mammography (CEDM) or angiomammography was developed. Objective: Assess whether CEDM is an effective method in the detection of breast cancer, as well as its reliability to rule out the presence of malignancy. Methods: Patients were recruited at the time of attendance to the service for breast percutaneous biopsy of lesions detected on previous mamography and /or ultrassound examinations, previously requested by their physicians. Those who made themselves available to participate in this study did sign the Informed Consent Form (ICF). Patients were submited to bilateral mammographic study in craniocaudal and mediolateral oblique incidences, obtaining low-energy images and recombined high-energy images. Low-energy images, equivalent to digital mammography, were described and classified according to the BI-RADS lexicon. The contrasted studies were described in order to comply with the recommendations of the current literature, observing that, until the present, there is no standardization by BI-RADS for contrast mammography reports. These studies were compared to histopathological findings of biopsies, the gold standard in this study. Results: This is an ongoing investigation. From September 2019 to October 2020, 180 patients underwent CEDM and percutaneous biopsies. 27 had invasive ductal carcinomas (IDC) and 10 ductal carcinomas in situ (DCIS). 26 of the 27 IDC cases and all of the DCIS cases had positive CEDM findings. Among DCIS cases, six had no abnormal enhancement on CEDM, but were evident on a 2D study. The observed sensitivity was 97%. Conclusions: These preliminary results demonstrated that CEDM is a highly sensitive method for breast cancer detection, including for non-invasive lesions. The study is still in progress and further data is needed to describe the benefits of CEDM in breast cancer detection.

Introduction The Standard of Care for early detection of breast cancer in asymptomatic women is screening mammography, which has limitations such as radiation exposure and lower sensitivity to detect cancer in women with high breast density or invasive carcinomas. TriNetra™-Breast is a blood test for the detection of breast cancer associated circulating tumor cells in blood. Previously, this test has been used in a study for breast cancer detection in India, where it has shown a sensitivity of 92.5%. It has since been granted the United States Food and Drug Administration (USFDA) Breakthrough Device Designation, attesting its potential to provide for improved detection of breast cancer. This prospective, observational, case-cohort study will confirm the clinical performance characteristics of the technology in the US population. Patients and Methods The primary endpoint of this study will be to determine the sensitivity and specificity of the test for breast cancer screening, using mammography and histopathology confirmed diagnosis (when relevant) as the reference methods. Women ≥40 years, with no prior diagnosis of any cancer and undergoing screening mammography for breast cancer will be eligible for participation in this study. 700 women, representing the diverse ethnic US population, will be enrolled. Cohort A will have 500 women with BI-RADS score of 1, 2, or 3. Among these 500 participants, the age categories of 40-49 years, 50-74 years and >74 years will have 100, 300 and 100 women respectively. Cohort B will have 100 women with suspicion of DCIS (without a suspicion of simultaneous invasive carcinoma) and 50 women each with BI-RADS score of 4 or 5. These study population numbers will ensure optimal representation of in-situ carcinoma, malignant and benign cases. Blood samples will be collected from the enrolled women for TriNetra™-Breast, within sixty (60) days of the screening mammogram. If biopsy is indicated, sample collection will be required prior to the procedure. The lab investigators will be blinded to the clinical information of all participants, including mammography and histopathology results, while the participants and clinical investigators will be blinded to the TriNetra™-Breast test results. The results of TriNetra™-Breast will be compared with the results of mammography and/or histopathology for performance estimation of the test. Study participants will be followed for clinical outcomes for maximum duration of 2 years. Citation Format: Ulka Vaishampayan, Darshana Patil, Wahida Rahman, Stephanie Patterson, Emilija Mitrikeska, Joe Dib. Clinical validation of “TriNetra™-Breast” test for breast cancer screening in a prospective, observational, case-cohort study [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 OT3-18-02.

Breast cancer positions as the most well-known threat and the main source of malignant growth-related morbidity and mortality throughout the world. It is apical of all new cancer incidences analyzed among females. However, machine learning algorithms have given rise to progress across different domains. There are various diagnostic methods available for cancer detection. However, cancer detection through histopathological images is considered to be more accurate. In this research, we have proposed the Stacked Generalized Ensemble (SGE) approach for breast cancer classification into Invasive Ductal Carcinoma+ and Invasive Ductal Carcinoma-. SGE is inspired by the stacking model which utilizes output predictions. Here, SGE uses six deep learning models as level-0 learner models or sub-models and Logistic regression is used as Level – 1 learner or meta – learner model. Invasive Ductal Carcinoma dataset for histopathology images is used for experimentation. The results of the proposed methodology have been compared and analyzed with existing machine learning and deep learning methods. The results demonstrate that the proposed methodology performed exponentially good in image classification in terms of accuracy, precision, recall, and F1 measure.

Presented here are the results of an investigation conducted to determine the effectiveness of deep learning (DL)‐based systems utilizing the power of transfer learning for detecting breast cancer in histopathological images. It is shown that DL models that are not specifically developed for breast cancer detection can be trained using transfer learning to effectively detect breast cancer in histopathological images. The outcome of the analysis enables the selection of the best DL architecture for detecting cancer with high accuracy. This should facilitate pathologists to achieve early diagnoses of breast cancer and administer appropriate treatment to the patient. The experimental work here used the BreaKHis database consisting of 7909 histopathological pictures from 82 clinical breast cancer patients. The strategy presented for DL training uses various image processing techniques for extracting various feature patterns. This is followed by applying transfer learning techniques in the deep convolutional networks like ResNet, ResNeXt, SENet, Dual Path Net, DenseNet, NASNet, and Wide ResNet. Comparison with recent literature shows that ResNext‐50, ResNext‐101, DPN131, DenseNet‐169 and NASNet‐A provide an accuracy of 99.8%, 99.5%, 99.675%, 99.725%, and 99.4%, respectively, and outperform previous studies.

Breast cancer is the second most dangerous disease for women after lung cancer. As in most diseases, an early detection and the corresponding treatment of breast cancer increase the survival rate of patients. An automated system for detection of breast cancer is required as manual techniques are time consuming and expensive. In this study, we have proposed a novel transfer learning model that can classify an ultrasound of the breast as either normal, benign, or malignant with high accuracy. The proposed method uses a modified ResNet50 model trained initially on ImageNet dataset and further on the breast cancer ultrasound dataset (BUSI). We have added custom layers at the head of our model which are able to extract features from ultrasound images. Using the model described in this paper, we have achieved 97.8% accuracy in detecting breast cancer, a recall of 97.68%, precision of 99.21% and 98.44% F1-score. This deep learning model can be implemented as a component of an existing medical diagnosis system or deployed as a stand-alone system. Using our model for breast cancer diagnosis can result in decreased diagnosis time compared to traditional means and hence ensure that patients receive an early treatment for their illness.

Breast cancer is the most common malignancy affecting women worldwide, and early diagnosis of breast cancer is the key to its successful and effective treatment. Traditional imaging techniques such as mammography and ultrasound are used to detect and configure breast abnormalities; unfortunately, these modalities have low sensitivity and specificity, particularly in young patients with dense breast tissue, breast implants, or post-surgical scar/architecture distortions. Therefore, breast magnetic resonance imaging (MRI) has been superior in the characterization and detection of breast cancer, especially that with invasive features. This review article explores the importance of breast MRI in the early detection of invasive breast cancer versus traditional tools, including mammography and ultrasound, while also analyzing the use of MRI as a screening tool for high-risk women. We will also discuss the different MRI features for invasive ductal carcinoma and lobular carcinoma and the role of breast MRI in the detection of ductal carcinoma in situ with a focus on the utilization of new techniques, including MR spectroscopy and diffusion-weighted imaging.

Breast cancer is the second biggest cause of cancer death among women in the United States. The most prevalent subtype of all breast cancers is invasive ductal carcinoma which is invasive ductal carcinoma. Traditional methods are hard to predict this cancer well. In this paper, we proposed a self-defined convolutional neural network (CNN) and employed a typical structure named ResNet for classification. Moreover, we compared the CNN model and the ResNet-50 model's ability in detecting breast cancer. Different parameters e.g. batch size were also changed in terms of different situations to find the difference of the performance. We built 4 sets of experiments for testing the collected datasets by testing the training accuracy and validation accuracy of the two models. The result showed that the accuracy of the ResNet-50 is higher than that of a self-defined CNN. Therefore, the new method should be applied in the process of the detection of breast cancer.

Introduction: The use of social media by healthcare professionals to create disease awareness has been increased worldwide. It is easily available and accessible to the larger number of populations and hence helps in bridging the gap between doctor and patient. Breast cancer is the leading cause of cancer among women in India and the main reason for higher mortality among them is the late presentation to the surgeons, owing to various myths and social and physical taboos. We conducted a study to assess the awareness of breast cancer among the women in general population and also to check the effectiveness of WhatsApp in improving women's knowledge regarding the risk reduction, myths, early detection and treatment of breast cancer. The study was a pre-post health educational study including 146 wom Methods: en. Multiple WhatsApp groups were created to share information regarding breast cancer awareness, early detection, risk factors, common myths, self-breast examination and treatment of breast cancer in the form of audio, video, text and images. The group was active for 4-5 days as an interactive session. Data were collected through google forms. We noticed a signicant improvement in the knowledge of women regard Results: ing the common myths regarding breast cancer and the symptoms of breast cancer. There was signicant increase in performance of self-breast examination from 14.4% to 84.6% (p- <0.001). Most of the women found the video of self-breast examination useful. The commonly Conclusion: used social networking site like WhatsApp can be used as a tool to assess the awareness of breast cancer among women. This also helped in exchange of the knowledge among women, which in turn helped in the risk reduction and early detection of breast cancer. This app is a useful tool that can be incorporated into health education strategies that focus on breast cancer.

Breast cancer is one of the most common cancers occurring in women and is the main cause of mortality among women aged 20 to 59. Detection of breast cancer in the early stage increases survival rate to 80%, and the rate decreases to around 40% if diagnosed in its later stage. There exists significant need to diagnose breast cancer in its early stage itself among women suspected with the disease. The hematoxylin and eosin (H&E)–stained histopathology tissue samples taken by pathologists from suspected breast areas are viewed with whole slide imaging (WSI) and can be stored as digital histopathology images. Manual analysis of histopathology images is time-consuming and leads to inter- and intraobserver variability. Hence, automated classification methods are in research focus. Breast tumors are classified as benign and malignant tumors, which are non-cancerous and cancerous tumors, respectively. Currently, deep learning through convolutional neural networks is an emerging technology and has proven to be successful for image segmentation and classification applications. These deep CNNs require large labeled datasets, which are lacking in the field of medicine, especially for breast cancer images. Also, high computational power, time, and memory are required if CNN is to be trained from scratch. To overcome this, transfer learning from natural images like ImageNet, comprising of 1.2 million natural images belonging to 1,000 classes, is used. Pretrained models like VGG16, ResNet50, Inception, AlexNet, and DenseNet are employed for transfer learning, and fine-tuning is to be performed for refinement. In the proposed method, four different network architectures, namely, AlexNet, VGG16, Inception v3, and DenseNet121, are used for classification of histopathological images using transfer learning. The top dense layers of these architectures are removed, and custom dense layers are added for classification. Also, the weights of the last two layers are fine-tuned with the histopathology input images dataset. The publicly available breast cancer hisopathology images dataset BreakHis is used for validation of the proposed method. The experimental results prove that transfer learning–based CNN gives better classification accuracy compared to networks learned from scratch. Also, it is inferred from the experimental results that the pretrained model of DenseNet121 gives the highest classification accuracy of 0.9520 compared to the other models and state-of-the-art methods.

Breast cancer detection from histopathological images has benefited significantly from advances in deep learning; however, many high-performing models rely on deep architectures and substantial computational resources, limiting their applicability in resource-constrained clinical settings. Recent studies have demonstrated that convolutional neural networks (CNNs) can effectively support invasive ductal carcinoma (IDC) classification, but often at the cost of increased training time and memory usage [1, 13].In this study, we propose BCDNet_Updated, an optimized lightweight CNN derived from the reference BCDNet architecture [1] for IDC detection. The proposed model introduces targeted architectural refinements aimed at improving classification performance while preserving computational efficiency. Experiments conducted on the Kaggle IDC dataset demonstrate that BCDNet_Updated achieves improved accuracy, recall, and F1-score compared to the reference BCDNet, while requiring significantly less GPU memory and comparable training time.While the evaluation is limited to a single dataset and does not include formal statistical significance testing or explainability analysis, the results indicate that careful optimization of lightweight CNN architectures can yield meaningful performance improvements without increasing computational cost. Future work will incorporate multi-dataset validation, statistical rigor through k-fold cross-validation, and explainability mechanisms (Grad-CAM, attention visualization) to support clinical adoption. This work contributes to ongoing efforts toward efficient deep learning models for histopathological image analysis and provides a foundation for future studies involving expanded validation and model interpretability.

Breast cancer remains one of the major global health challenges, the Human Epidermal Receptor-2 (HER-2) subtype is characterized by aggressive tumor growth due to overexpression of the HER2-positive. Early detection of HER2-positive breast cancer is crucial for improving patient’s outcomes and reducing the risk of metastatic disease. Circulating Tumor DNA is DNA fragments which are released by cancer cells into the bloodstream, and has emerged as a promising non-invasive biomarker for early cancer detection. This review will explore the role of ctDNA in the early detection of HER2-positive breast cancer and its role in prognosis. Studies have shown the high relevance of ctDNA as a diagnostic biomarker, particularly in early-stage disease. Multiple studies have evaluated the role of ctDNA in breast cancer patients, which results consistently showed a strong relation between ctDNA detection and early-stage disease. Additionally, research has indicated that high ctDNA levels are related with a higher risk of recurrence and poorer survival outcomes. The non-invasive nature of ctDNA testing offers several advantages compared to traditional methods, including the ability of monitoring disease progression and treatment response without any need for invasive procedures. Early detection of HER2-positive breast cancer through ctDNA analysis can potentially improve patient outcomes by enabling timely intervention and targeted therapies. By identifying patients with high risk and monitoring disease status, ctDNA testing may contribute to more effective management of HER2-positive breast cancer. While ctDNA analysis shows great promise, it is also important to acknowledge the current limitations and challenges associated with its clinical application. Issues such as standardization of detection methods, sensitivity, and specificity need to be addressed to ensure the widespread adoption of ctDNA testing in routine clinical practice. In conclusion, ctDNA represents a promising biomarker for the early detection and management of HER2-positive breast cancer. Further research and advancements in ctDNA analysis are essential to completely realize its potential and improve patient outcomes.