We propose a lightweight computer-aided diagnosis (CAD) framework that fuses four sub-band discrete wavelet transform gray-level co-occurrence matrix (DWT–GLCM) texture features with fine-tuned ResNet-50 embeddings under a strict, patient-level, leak-free evaluation protocol. Experiments were conducted on two public datasets: mammographic image analysis society (MIAS) (normal vs. abnormal) and curated breast imaging subset of the digital database for screening mammography (CBIS-DDSM) (benign vs. malignant). Five-fold cross-validation (CV) was confined to the training portion, operating thresholds were fixed on the validation split to target high recall, and the held-out test set was evaluated once. Performance was assessed using accuracy, F1-score, receiver operating characteristic (ROC)-area under the curve (AUC) with bootstrap 95% confidence intervals (CI), precision-recall (PR)-AUC, and calibration metrics (Brier score, expected calibration error). The proposed fusion model achieved ROC-AUC on MIAS (0.992) and strong performance on CBIS-DDSM (0.896), with consistent PR characteristics. Calibration analysis indicated reliable probability estimates and clinically interpretable decisions at a 95% sensitivity operating point. Ablation experiments revealed substantial gains over texture-only baselines and parity with convolutional neural network (CNN)-only models, highlighting fusion as a simple yet well-calibrated alternative for screening-oriented workflows. This study underscores the necessity of patient-level evaluation, explicit operating-point selection, and calibration reporting to ensure clinically meaningful CAD performance in mammography.

Breast cancer is one of the major problems affecting the breast that is very commonly detected in females, and it requires efficient and precise diagnosis for enhanced stages of survival. To achieve efficient, precise, and contactless diagnostic processes, it is proposed that the Radio Thermal Dual-Imaging Fusion Framework be utilized in combination with the structural information obtained from mammogram images, along with thermal information obtained from infrared images of thermograms of the breasts. This paper proposes a conceptual design that makes use of an available set of mammogram images taken from digital image databases, such as The Mammographic Image Analysis Society Database (MIAS), standard sets of samples from the curated Breast Imaging Subset of the Digital Database for Screening Mammography (CBIS-DDSM), as well as samples from anonymized patients in recent routine clinical screenings, similar to recent diagnostic studies that utilized mammography. The samples for thermal images were obtained from infrared images of thermograms of breasts, which are still very much in use in today’s breast cancer diagnostic research. The proposed conceptual design for diagnosis will employ the Super-Fast Recurrent Convolutional Neural Network (SFRCNN) architecture, which will be a ResNet-50 based architecture for the extraction of thermal images and standard images, as mentioned above. The preprocessing for sampling the modalities will be done using grayscale normalization of the standard mammogram images as well as standard thermal mapping for the infrared images. From the results of the proposed conceptual design, it has been identified that the proposed dual-imaging modality accuracy of 84.75% will be obtained by using the standard mammogram images and the standard infrared images, representing a significant improvement over standard image processing, as an improvement in accuracy of 3% is expected from the proposed standard approaches within the benchmark design for the Convolutional Neural Network model due to the proposed dual-imaging modality technique, along with a proposed diagnosis accuracy for assessing a sensitivity of 98.04% by the dual-imaging modalities of the standard infrared images.

To increase patient survival rates, breast cancer must be detected early and, most importantly, accurately. Although mammography analysis using deep learning has advanced, generalizability is still hampered by issues including dataset variability, class imbalance, and inter-class similarity. Using pre-trained DenseNet121, InceptionV3, and EfficientNetB0 models, this study presents a unique and robust ensemble-based deep learning architecture that was independently refined on mammography images from the INbreast and Mammographic Image Analysis Society (MIAS) datasets. The pipeline incorporates Contrast Limited Adaptive Histogram Equalization (CLAHE) for preprocessing, employs focal loss to mitigate class imbalance, and utilizes a two-stage refinement strategy in which misclassified samples are reintroduced for model retraining. This approach achieves significant improvements, reaching up to 99% and 97% classification accuracy on the INbreast and MIAS datasets, respectively. The ensemble model with misclassification refinement demonstrates notable robustness and generalization capability. The same pipeline was applied independently to both datasets to perform cross-dataset validation, confirming its adaptability to diverse mammographic characteristics. Accuracy, precision, recall, F1-score, and misclassification analysis were employed to evaluate the method, making it suitable for real-world clinical applications. The proposed framework holds strong potential for deployment in clinical decision support systems and lays a solid foundation for future research on small and heterogeneous medical datasets.

Breast cancer is the most frequently diagnosed and leading cause of cancer-related mortality among women worldwide. The danger of this disease is due to its asymptomatic nature in the early stages, thereby underscoring the importance of early detection. Mammography, a specialized X-ray imaging technique for breast examination, has been pivotal in facilitating early detection and reducing mortality rates. In recent years, artificial intelligence (AI) has gained substantial popularity across various fields, including medicine. Numerous studies have leveraged AI techniques, particularly convolutional neural networks (CNNs) and You Only Look Once (YOLO)-based models, for medical image detection and classification. However, the predictions of such AI models often lack transparency and explainability, resulting in low trustworthiness. This study aims to address this gap by investigating three state-of-the-art versions of the YOLO algorithm-YOLO version 9 (YOLOv9), YOLO version 10 (YOLOv10), and YOLO version 11 (YOLO11)-trained on breast cancer imaging datasets, specifically the INbreast and Mammographic Image Analysis Society (MIAS) databases. Additionally, to address the challenges posed by the lack of explainability and transparency, we integrate seven explainable artificial intelligence (XAI) methods: Grad-CAM, Grad-CAM++, Eigen-CAM, EigenGrad-CAM, XGrad-CAM, LayerCAM, and HiResCAM. This study utilized two publicly available breast cancer image databases: INbreast: toward a Full-field Digital Mammographic Database and the MIAS dataset. Preprocessing steps were applied to standardize all images in accordance with the input requirements of the YOLO architecture, as these datasets were used to train the three most recent versions of YOLO. The YOLO model demonstrating the highest performance-measured by mean average precision (mAP), precision, and recall-was selected for integration with seven different XAI methods. The performance of each XAI technique was evaluated both qualitatively through visual inspection and quantitatively using several metrics, including matching ground truth (mGT), Pearson correlation coefficient (PCC), precision, recall, and root mean square error (RMSE). These methodologies were employed to interpret and visualize the "black box" decision-making processes of the top-performing YOLO model. Based on our experimental findings, YOLO11 outperformed YOLOv9 (mAP 0.868) and YOLOv10 (mAP 0.926), achieving the highest mAP of 0.935, with classification accuracies of 95% for benign and 80% for malignant cases. Among the evaluated XAI techniques, HiResCAM provided the most effective visual explanations, attaining the highest mGT score of 0.49, surpassing EigenGrad-CAM (0.45) and LayerCAM (0.42) in both visual and quantitative evaluations. The integration of YOLO11 with HiResCAM offers a robust solution that combines high detection accuracy with improved model interpretability. This approach not only enhances user trustworthiness by revealing decision-making patterns and limitations but also provide insights into the weaknesses of the model, enabling developers to refine and improve AI performance further.

Background and Objective : The mortality rates due to breast cancer have been constantly growing and still represent one of the most common malignancies leading to death in females globally. Early and accurate detection is crucial to improve the survival rate. Recent deep learning advancements in artificial intelligence have opened a wide new avenue for further improving the results of computer-aided diagnosis. Vision transformers with their attention mechanism are among the recent promising ones, offering much-improved results for different image analysis applications, including mammography. Methods : This study investigates the application of vision transformers and attention mechanisms for mammography image categorization. In this work, we used three publicly available datasets like the Mammographic Image Analysis Society (MIAS), Curated Breast Imaging Subset of DDSM (CBIS-DDSM), and INbreast. In the preprocessing of data, augmentation is used to enhance the generalization capabilities of models, and we have applied Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve the contrast of images, especially in situations characterized by uneven lighting or low contrast levels. Results : The proposed approach demonstrated superior performance compared to traditional convolutional neural network (CNN)-based methods. In the evaluation of this vision transformer, we have obtained an accuracy of 0.99, an AUC of 0.99 and an F1 score of 0.98. Conclusion : Vision transformers and attention mechanisms have great potential to boost the detection of breast cancer using CAD systems. The findings accentuate their capability to improve the precision and reliability of mammography analysis, enabling early diagnosis and minimizing false positives and false negatives in clinical practice. The research emphasizes the need to embrace these new technologies to enhance patient outcomes and streamline healthcare resources.

Breast Cancer (BC), characterized by the uncontrolled proliferation of breast cells, remains the most prevalent and life-threatening cancer affecting women. Early detection significantly increases the chances of survival by enabling timely medical intervention. Numerous methods have been proposed for breast cancer detection; however, limitations in diagnostic accuracy and efficiency persist. To address these challenges, this study introduces a robust deep-learning framework leveraging fine-tuned EfficientNet-B3 for the detection and classification of breast tumors. The methodology employs segmentation techniques to accurately delineate the affected breast tumor regions, reducing training complexity while enhancing classification precision. The model was trained and evaluated using the Mammographic Image Analysis Society (MIAS) dataset, incorporating critical preprocessing steps to optimize image quality and feature extraction. Fine-tuning of EfficientNet-B3 was carried out to adapt the pre-trained network to mammogram-specific features, with hyperparameters optimized for this domain. Performance was assessed using three primary evaluation metrics: accuracy, specificity, and sensitivity. Experimental results demonstrate that the proposed CNN-EfficientNet-B3 framework achieves superior performance compared to conventional approaches, with specificity, accuracy, and sensitivity rates of 97.12%, 96.50%, and 96.43%, respectively. These findings highlight the potential of the proposed method to significantly enhance breast cancer detection and classification, paving the way for more effective clinical applications.

Introduction: When considering cancer mortality rates in general, Breast Cancer (BC) is a major contributor among females. Patients’ chances of survival increase when BC is detected early and treated with the appropriate treatment at the right time. There is strong evidence that mammography, when used as a screening tool, can detect BC at an early stage. Mammography is a diagnostic tool that uses low-dose X-rays to visualise the breast and evaluate its anatomy. For screening purposes, it is currently the preferred method. The present study employs deep learning models trained using Transfer Learning (TL) techniques. Aim: To automate the process of BC diagnosis in mammograms. The main goal of this approach is to simplify the process of early detection and diagnosis of BC for healthcare practitioners. Materials and Methods: The dataset obtained from the Mammographic Image Analysis Society (MIAS) was categorised into three distinct categories: benign, malignant and normal. The initial MIAS dataset underwent several preprocessing techniques, including noise reduction, breast image contrast enhancement, non breast region deletion and malignant lesion identification, before analysis. An intricately designed fully connected classifier complements pretrained Convolutional Neural Network (CNN) architectures like ResNet50 and VGG16 in the proposed model. Results: The VGG16 model performed admirably, achieving an Area Under the Curve (AUC) of 0.950 and an accuracy rate of 96.00%. In addition, it displayed an outstanding F-score of 97%, along with high sensitivity, specificity and accuracy. These outcomes are significantly better compared to the other methods. Conclusion: The model’s enhanced capabilities for early-stage cancer detection could improve patient outcomes and reduce mortality rates. Furthermore, new tools can ease the workload for radiologists, leading to more standardised and efficient diagnostic procedures.

Abstract Breast cancer remains one of the most prevalent and life-threatening diseases affecting women worldwide. Early detection is crucial for improving survival rates and treatment outcomes. This study proposes an advanced feature extraction method for classifying mammogram masses by combining multi-scale multi-orientation (MSMO) Gabor wavelets and gray-level co-occurrence matrix (GLCM) statistical features. MSMO Gabor filters extract detailed texture information across multiple scales and orientations, while GLCM captures statistical spatial relationships between pixel intensities. A feature selection process refines these features, enhancing classification accuracy. Experiments using Mammographic Image Analysis Society (MIAS) and Digital Database for Screening Mammography (DDSM) datasets validate the approach with machine learning classifiers, including random forest (RF), decision tree (DT), support vector machine (SVM), k-nearest neighbors (k-NN), and deep neural network (DNN). RF outperformed other models and achieved 96.64% accuracy on MIAS dataset and 95.90% on DDSM dataset. Our approach shows the efficacy of optimally combining MSMO Gabor and GLCM features to advance computer-aided diagnosis systems for early and precise breast cancer detection.

In the recent years, the field of breast cancer research is using deep learning techniques to mitigate the problems of false positive and false negative cases caused by the breast cancer diagnosis done by the radiologist. Therefore, in this research we propose deep CNN (convolution neural networks) for classifying the mammogram image as cancerous or non-cancerous. However, it is observed that, merely using deep learning techniques also has some limitations such as, uncertainty in breast mass classification on the dense breast mammograms. Thus, in this research the images are preprocessed and segmented prior classification. In this research images are preprocessed using Rolling Ball algorithm for background removal and then compared CLAHE and Unsharp masking for improving image contrast and visibility of the images. The proposed work segmented MIAS preprocessed images using k-means algorithm. These enhanced images are provided to Deep CNN (Convolution Neural Networks) and its features extracted for classifying images as benign, malignant or normal. To improve the CNN's structure and lessen overfitting, dropout and zero-padding are employed. The proposed work is tested on mammography images from the databases of the Mammographic Image Analysis Society (mini-MIAS), Breast Cancer Digital Repository (BCDR). The proposed system has, Accuracy:0.89%, Precision: 0.91%, Recall: 0.90%, F1-score 0.88% and for the MIAS dataset, respectively.

Breast cancer (BC) is one of the most fatal diseases that have a profound impact on women. If the cancer is identified earlier, the proper treatment will be provided to the patients to decrease the possibility of death. Mammography is a widely used imaging modality to detect BC earlier, providing valuable information to radiologists to offer better treatment plans and outcomes. This article proposes an efficient BC detection system from mammographic images using a hyperparameter-tuned gated recurrent unit (HTGRU) with attention included in a pre-trained model. The system includes the following steps: preprocessing, segmentation, feature extraction, and classification. The proposed system performs preprocessing using Gaussian filtering and contrast-limited adaptive histogram equalization (CLAHE) for noise removal and contrast enhancement. The data augmentation is performed on the preprocessed dataset to balance the data samples of the benign and malignant classes that prevents the network form biased results. After that, a deviation theory-based fuzzy c-means (DTFCM) algorithm is utilized to segment the tumor regions from the preprocessed image. Then, the most discriminant features are extracted from the segmented tumor regions using a normalization-based attention module incorporated in the capsule network (NAMCN). Finally, HTGRU is used for classification, classifying the data into benign, malignant, and normal. The system is evaluated by the Mammographic Image Analysis Society (MIAS) and curated breast imaging subset of a digital database for screening mammography (CBIS-DDSM) datasets, and the outcomes demonstrate the proposed method’s superiority over existing methods by achieving higher detection accuracy and lower false positive rates.

Noise and artefacts in mammogram images can obscure important indicators of microcalcifications, complicating accurate diagnosis. While traditional spatial filters can reduce noise and are effective to some extent, they often fail to enhance features crucial for classification. This study uses persistent homology (PH) to evaluate and improve the classification performance of various spatial filters on mammogram images. The evaluation process involves converting filtered images into persistence diagrams (PDs) to capture topological features. These diagrams are then vectorised into PH features for classification using a neural network classifier. This study also examines further filtering of PDs from filtered images to enhance classification performance. Using the Digital Database for Screening Mammography (DDSM) and Mammographic Image Analysis Society (MIAS) datasets, we evaluate Median, Wiener, Gaussian, and Bilateral filters alone and integrate them with PH-based filtering. Results show significant classification improvements, with Wiener filters achieving 96.33% accuracy on the DDSM dataset (up from 57.38%) and Gaussian filters reaching 85.33% on the MIAS dataset (up from 73.33%). These findings demonstrate the potential of PH-based filters to enhance diagnostic accuracy in breast cancer detection by refining topological features and effectively reducing noise.

Among all new cancer cases diagnosed, breast cancer has been leading in count, followed by prostate and lung cancer.Breast cancer also has the highest chances of getting cured, if it gets early diagnosis, thus increasing the lives of not only women but also the minority of males.For the same, the Deep Learning algorithms with transfer learning models are utilized, already trained with ImageNet database, and partially training them on the small mammography images database and thus help to diagnose it without the need for large datasets or tissue analysis (biopsy).The pre-trained convolution neural network models of VGG-16, VGG-19, ResNet50 and Inception V3 are worked as Deep Transfer Learning on two databases: the Mammography Image Analysis Society (MIAS) database containing 321 images, and Chinese Mammography Database (CMMD) containing 3744 mammography, of which 2000 images are used for learning.The evaluation of the model is based upon the parameters of accuracy, precision, recall, and F1-score.For MIAS Database, VGG 19 model showed better results, with accuracy being 98.44%, and precision, recall and F1 score being 0.99 each.For CMMD, VGG16 showed better results, with accuracy being 99.50%, precision being 1.0, recall being 0.99, and F1 score of 0.99.

Fine-tuning pre-trained networks with emphasis on image segmentation: A multi-network approach for enhanced breast cancer detection

ABSTRACT Breast cancer is cancer that forms in the cells of the breasts and is a severe health issue that affects many people around the world, especially since it is the most deadly cancer in women. By finding it early and using new treatments, patients can overcome this challenge and get back to a healthier life. This study proposed a procedure to fine-tune the Convolutional Neural Networks (CNN) model with data preprocessing and augmentation in classifying mammogram images called the Hybrid Mammogram Classification and Detection Pipeline (HMCaD). After using CNN for classification because it brings higher confidence in classifying tasks, the YOLOv8 has been applied for localization subtask to detect abnormal positions with predicted bounding boxes. The database is provided by the Mammographic Image Analysis Society (MIAS) and is protected by the United Kingdom. It comprises 330 samples, including 79 benign, 54 malignant, and 207 normal images. As a result, the classification in our model based on the custom EfficientNetB3 model and seam carving technique received a great validation accuracy, test accuracy, and F1 score throughout three scenarios at 96.73%, 97.59%, and 97.58%, respectively. Furthermore, the area under the Receiver Operating Characteristic (ROC) curve also has a surprise result of 0.96 (i.e. AUC = 0.96 ). Moreover, YOLOv8 for detecting abnormal positions in our study achieved 83.22% in Intersection over Union (IoU). This led to the research reaching good results in classifying and detecting breast cancer by considering several performance metrics.

Breast cancer has become a major worldwide health issue, accounting for a large portion of the mortality rate among women. As a result, the need for early detection techniques to enhance prognosis is increasing. Many techniques are being used to detect breast cancer early, and treatment outcomes are frequently favorable when the disease is detected early. Mammography is a commonly used and very successful method for identifying breast cancer among these modalities. Through additional image processing operations like resizing and normalizing, this technology allows the detection of malignant spots from mammography pictures of the affected area. The goal of our research is to improve breast cancer detection and diagnosis speed and accuracy. In this study, we investigate the use of deep learning methods, particularly the visual geometry group (VGG16) and ResNet50 models, for mammography-based breast cancer detection. We assess the performance of the VGG16 and ResNet50 models by training and testing on a mammogram dataset that consists of 322 images from the mammographic image analysis society (MIAS) dataset. The suggested models aim to classify these images into normal, benign, and malignant groupings. Our results show better performance when compared to existing approaches. The proposed methods VGG16 and ResNet50 show promising results, achieving a classification accuracy of 91.23% and 99.01% respectively.

The early detection of breast cancer is essential for improving treatment outcomes, and recent advancements in artificial intelligence (AI), combined with image processing techniques, have shown great potential in enhancing diagnostic accuracy. This study explores the effects of various image processing methods and AI models on the performance of early breast cancer diagnostic systems. By focusing on techniques such as Wiener filtering and total variation filtering, we aim to improve image quality and diagnostic precision. The novelty of this study lies in the comprehensive evaluation of these techniques across multiple medical imaging datasets, including a DCE-MRI dataset for breast-tumor image segmentation and classification (BreastDM) and the Breast Ultrasound Image (BUSI), Mammographic Image Analysis Society (MIAS), Breast Cancer Histopathological Image (BreakHis), and Digital Database for Screening Mammography (DDSM) datasets. The integration of advanced AI models, such as the vision transformer (ViT) and the U-KAN model—a U-Net structure combined with Kolmogorov–Arnold Networks (KANs)—is another key aspect, offering new insights into the efficacy of these approaches in different imaging contexts. Experiments revealed that Wiener filtering significantly improved image quality, achieving a peak signal-to-noise ratio (PSNR) of 23.06 dB and a structural similarity index measure (SSIM) of 0.79 using the BreastDM dataset and a PSNR of 20.09 dB with an SSIM of 0.35 using the BUSI dataset. When combined filtering techniques were applied, the results varied, with the MIAS dataset showing a decrease in SSIM and an increase in the mean squared error (MSE), while the BUSI dataset exhibited enhanced perceptual quality and structural preservation. The vision transformer (ViT) framework excelled in processing complex image data, particularly with the BreastDM and BUSI datasets. Notably, the Wiener filter using the BreastDM dataset resulted in an accuracy of 96.9% and a recall of 96.7%, while the combined filtering approach further enhanced these metrics to 99.3% accuracy and 98.3% recall. In the BUSI dataset, the Wiener filter achieved an accuracy of 98.0% and a specificity of 98.5%. Additionally, the U-KAN model demonstrated superior performance in breast cancer lesion segmentation, outperforming traditional models like U-Net and U-Net++ across datasets, with an accuracy of 93.3% and a sensitivity of 97.4% in the BUSI dataset. These findings highlight the importance of dataset-specific preprocessing techniques and the potential of advanced AI models like ViT and U-KAN to significantly improve the accuracy of early breast cancer diagnostics.

Medical image classification (IC) is a method for categorizing images according to the appropriate pathological stage. It is a crucial stage in computer-aided diagnosis (CAD) systems, which were created to help radiologists with reading and analyzing medical images as well as with the early detection of tumors and other disorders. The use of convolutional neural network (CNN) models in the medical industry has recently increased, and they achieve great results at IC, particularly in terms of high performance and robustness. The proposed method uses pre-trained models such as Dense Convolutional Network (DenseNet)-121 and Visual Geometry Group (VGG)-16 as feature extractor networks, bidirectional long short-term memory (BiLSTM) layers for temporal feature extraction, and the Support Vector Machine (SVM) and Random Forest (RF) algorithms to perform classification. For improved performance, the selected pre-trained CNN hyperparameters have been optimized using a modified grey wolf optimization method. The experimental analysis for the presented model on the Mammographic Image Analysis Society (MIAS) dataset shows that the VGG16 model is powerful for BC classification with overall accuracy, sensitivity, specificity, precision, and area under the ROC curve (AUC) of 99.86%, 99.9%, 99.7%, 97.1%, and 1.0, respectively, on the MIAS dataset and 99.4%, 99.03%, 99.2%, 97.4%, and 1.0, respectively, on the INbreast dataset.

Breast cancer has become the leading cause of cancer mortality among women worldwide. The timely diagnosis of such cancer is always in demand among researchers. This research pours light on improving the design of computer-aided detection (CAD) for earlier breast cancer classification. Meanwhile, the design of CAD tools using deep learning is becoming popular and robust in biomedical classification systems. However, deep learning gives inadequate performance when used for multilabel classification problems, especially if the dataset has an uneven distribution of output targets. And this problem is prevalent in publicly available breast cancer datasets. To overcome this, the paper integrates the learning and discrimination ability of multiple convolution neural networks such as VGG16, VGG19, ResNet50, and DenseNet121 architectures for breast cancer classification. Accordingly, the approach of fusion of hybrid deep features (FHDF) is proposed to capture more potential information and attain improved classification performance. This way, the research utilizes digital mammogram images for earlier breast tumor detection. The proposed approach is evaluated on three public breast cancer datasets: mammographic image analysis society (MIAS), curated breast imaging subset of digital database for screening mammography (CBIS-DDSM), and INbreast databases. The attained results are then compared with base convolutional neural networks (CNN) architectures and the late fusion approach. For MIAS, CBIS-DDSM, and INbreast datasets, the proposed FHDF approach provides maximum performance of 98.706%, 97.734%, and 98.834% of accuracy in classifying three classes of breast cancer severities.

Breast cancer remains a significant global health challenge, necessitating advancements in screening and diagnostic methods for its early detection and treatment. This review explores the role of open-access mammography databases in facilitating research and development in the field of breast cancer detection, particularly through the integration of artificial intelligence techniques such as machine learning and deep learning. In this review, we highlight the open-access databases, including the Digital Database for Screening Mammography (DDSM), the Curated Breast Imaging Subset of DDSM (CBIS-DDSM), Mini-DDSM, INbreast, Mammographic Image Analysis Society Dataset (MIAS), and the China Mammography and Mastopathy Dataset (CMMD). Each database was analyzed in terms of its composition, features, limitations, and contributions to breast cancer research. In addition, we highlight the importance of open-access databases in enabling collaborative research, improving algorithm development, and enhancing the accuracy and efficiency of breast cancer detection methods computer-aided diagnosis.

Mammography is the modality of choice for the early detection of breast cancer. Deep learning, using convolutional neural networks (CNNs) specifically, have achieved extraordinary results in the classification of diseases, including breast cancer, on imaging. The images used to train a CNN varies based on several factors, such as imaging technique, imaging equipment, and study population; these factors significantly affect the accuracy of the CNN models. The aim of this study was to develop a novel CNN for the classification of mammograms as benign or malignant and to compare its utility to that of popular pre-trained CNNs in the literature using transfer learning. All CNNs were trained to detect breast cancer on mammograms using mammograms from a created database of Mexican women (MAMMOMX-PABIOM) and from a public database of UK women (MIAS). A database (MAMMOMX-PABIOM) was built comprising 1,070 mammography images of 235 Mexican patients from 4 hospitals in Mexico. The study also used mammographic images from the Mammographic Image Analysis Society (MIAS) public database, which comprises mammography images from the UK National Breast Screening Programme. A novel CNN was developed and trained based on different configurations of training data; the accuracy of the models resulting from the novel CNN were compared with models resulting from more advanced pre-trained CNNs (DenseNet121, MobileNetV2, ResNet 50, VGG16) which were built using transfer learning. Of the models resulting from pre-trained CNNs using transfer learning, the model based on MobileNetV2 and training data from the MAMMOMX-PABIOM database achieved the highest validation accuracy of 70.10%. In comparison, the novel CNN, when trained with the data configuration A6, which comprises data from both the MAMMOMX-PABIOM database and the MIAS database, produced a much higher accuracy of 99.14%. Although transfer learning is a widely used technique when training, data is scarce. The novel CNN produced much higher accuracy values across all configurations of training data compared to the accuracy values of pre-trained CNNs using transfer learning. In addition, this study addresses the gap in that neither a national database of mammograms of Mexican women exists, nor a deep learning tool for the classification of mammograms as benign or malignant that is focused on this population.