Automated Breast Cancer Detection using Hybrid CNN Models on Histopathological Images

Deep Learning-Based Automated Skin Cancer Detection Using Dermoscopic Images

Breast Cancer Detection Using Transfer Learning Contribution

: Breast cancer remains one of the most critical health concerns for women, and early detection plays a major role in improving survival rates. With the rapid growth of deep learning techniques, computer-based systems have become increasingly effective at analyzing and classifying medical images. In this work, deep learning particularly transfer learning is used to identify breast cancer from histopathological images. Transfer learning allows us to take advantage of powerful models that have already been trained on vast image datasets and adapt them to our specific task. For this study, three well-known pre-trained convolutional neural network (CNN) architectures InceptionV3, EfficientNet, and ResNet50 were evaluated to determine which performs best for breast cancer classification. Each model was fine-tuned using our collection of breast tissue images. During experimentation, we noticed that some models had difficulty learning the necessary patterns for correct classification. Among them, ResNet50 stood out by producing more reliable and accurate predictions. The workflow followed several essential stages: images were preprocessed and resized according to each model’s input requirements, and data augmentation was applied to increase variety and prevent overfitt ing. After preparing the dataset, the models were trained and assessed using standard evaluation metrics, helping us compare their strengths and weaknesses. The results clearly indicate that the choice of architecture has a major impact on performance, esp ecially in medical image analysis. ResNet50’s deeper network and skip connections enabled it to capture fine-grained features within the images, leading to improved accuracy. This demonstrates the strong potential of deep learning techniques in healthcare. Beyond building an effective classifier, this project also emphasizes how artificial intelligence can support medical professionals in diagnosis. A dependable system can minimize human error, assist doctors in decision-making, and accelerate the detection process particularly in regions where experienced pathologists may be scarce. Overall, this work shows that transfer learning is a practical and efficient method for breast cancer detection and could contribute to better patient outcomes in the future.

A Convolutional Neural Network (CNN)- based Approach for Lung Cancer Detection Leveraging Classical Edge Detection Techniques: An AI-based Retrospective Study

Introduction: Lung cancer remains one of the leading causes of cancer-related mortality worldwide. Despite advances in diagnostic imaging, histopathological assessment continues to be the gold standard for accurate classification of lung cancer subtypes. However, manual interpretation of histopathological slides is time-consuming and susceptible to inter-observer variability, underscoring the need for efficient Computer-Aided Diagnostic (CAD) systems. Aim: To develop a Convolutional Neural Network (CNN)- based model for automated classification of lung carcinoma histopathological images by integrating classical gradientbased edge detection techniques to enhance image quality and improve classification performance. Materials and Methods: This retrospective analytical study was conducted in the ICMR-funded laboratory at Brainware University, Kolkata, West Bengal, India. The project commenced in February 2024 and is proposed to conclude in February 2027. To date, it has yielded a dataset of 11,580 histopathological images of lung tissue, categorised as Lung Adenocarcinoma (LUAD), Lung Squamous Cell Carcinoma (LSCC), and Non Malignant (BNM). The present work represents a split study and forms part of a larger ICMR-funded project spanning three years. This component focuses on LUAD, LSCC, and non-malignant lung tissue, thereby encompassing the two most common Non Small Cell Lung Cancer (NSCLC) subtypes (which together account for approximately 85% of lung cancers) along with a normal reference class. Edge enhancement was performed using Sobel and Scharr operators to emphasise structural boundaries. The processed images were used to train a custom-designed CNN architecture comprising three convolutional blocks, each followed by pooling layers, and fully connected layers for classification. The model was trained for 100 epochs with a batch size of 32 and evaluated using accuracy, precision, sensitivity, specificity, and F1-score as performance metrics. Results: Sobel-based preprocessing yielded superior model performance compared with the Scharr method. The proposed CNN trained on Sobel-enhanced images achieved an overall classification accuracy of 93.55%, outperforming the 81.33% accuracy obtained with Scharr-based preprocessing. The Sobelenhanced model also demonstrated stronger generalisation, with AUC values of 0.98, 1.00, and 0.98 for the LUAD, BNM, and LSCC classes, respectively. Conclusion: The integration of edge detection with CNN-based analysis effectively enhances the classification of lung cancer histopathological images. The Sobel operator notably improved model accuracy, precision, and generalisation, suggesting its potential suitability for highlighting morphological boundaries in tissue sections. Future work will extend this framework by exploring additional edge detection methods, implementing stain normalisation, and validating performance on external datasets to further optimise model robustness and clinical applicability.

Deep Learning-Based Survival Prediction for Breast Cancer Using Whole Slide Histopathology Images

Breast cancer is currently one of the leading malignancies and mortality rates among women globally, creating an urgent need for accurate and efficient automated diagnostic tools. This study proposes and systematically compares convolutional neural network (CNN) architectures, including VG16, ResNet18, ResNet50, DenseNet121, EfficientNet-B0, and Swin-Transformer, applied to whole slide image (WSI) histopathology data for breast cancer prediction and classification. The models were trained and evaluated on the same WSI dataset with consistent image preprocessing techniques, allowing for objective comparison of performance. Experimental results showed that the performance of the models varied depending on the training strategy. In the non-fine-tune setting, ResNet18 achieved the best results with a C-index of 0.6587 and a Mean time-dependent AUC of 0.7317. When performing fine-tuning, ResNet50 outperformed the other architectures with a C-index of 0.6877 and a Mean AUC of 0.7738. Meanwhile, in the non-fine-tuned setup with combined clinical testing, Swin-Transformer achieved the highest performance with a C-index of 0.6793 and a Mean AUC of 0.7619.

A CNN–Transformer Hybrid Model for Early Breast Cancer Detection from Mammograms

Early-stage identification of breast cancer is imperative in decreasing the cancer-related mortality rate. However, the interpretation of mammograms is difficult as the contrast is low, and lesions of interests may be subtle. This study proposes a hybrid model which combines Convolutional Neural Networks (CNN) and Transformers, to provide an accurate and timely mechanism for the detection of breast cancer. The hybrid model captures the CNN’s ability to identify local features (e.g., edges and textures) and the Transformer’s ability to identify global features and context through self-attention. The model is implemented and validated on the MIAS and CBIS-DDSM datasets. The proposed method utilizes CLAHE for noise and contrast enhancement and captures features more effectively. The hybrid model also provides an improvement in Confirmed Contrast Enhancement (CCE) of 3.2 and greater entropy values when compared to baseline CNN models, ResNet and DenseNet. The results of the study indicate that the integration of CNN and Transformer, improves the reliability of the diagnosis, and supports the use of the model for the further development of a Real-time Clinical Decision Support (CDS) System for the early detection of breast cancer.

Deep Learning-Based Segmentation, Classification, and Explainable Diagnosis of Breast Cancer Lesions

Breast cancer remains one of the most prevalent malignancies affecting women worldwide, and early, accurate detection through medical imaging is central to improving survival outcomes. This paper presents an end-to-end computer-aided diagnosis (CAD) framework that performs lesion segmentation, region-of-interest (ROI) extraction, tumor stage estimation, benign/malignant classification, and visual explainability across two complementary imaging modalities: digital mammography, represented by the Curated Breast Imaging Subset of the Digital Database for Screening Mammography (CBIS-DDSM), and breast ultrasound, represented by the Breast Ultrasound Images (BUSI) dataset. For each modality, a modality-specific U-Net architecture was trained to generate binary lesion masks, from which tumor coverage percentage and a rule-based disease stage were derived. Three classifier backbones were benchmarked per modality — ResNet50, VGG16, and a custom convolutional neural network (CNN) for CBIS-DDSM (benign vs. malignant), and ResNet50, EfficientNet-B0, and a custom CNN for BUSI (benign vs. malignant vs. normal) — with Gradient-weighted Class Activation Mapping (Grad-CAM) applied to the strongest classifiers for interpretability. The trained pipelines were deployed in a Django-based web application allowing end users to upload an image and receive segmentation, ROI, stage, classification, confidence, and Grad-CAM outputs in real time. Results show a pronounced modality gap: the U-Net segmentation network achieved a validation Dice coefficient of 0.462 on BUSI versus 0.247 on CBIS-DDSM, and the best classifier reached 84.07% test accuracy on BUSI (ResNet50) compared with 61.86% on CBIS-DDSM (ResNet50), with markedly lower malignant-class recall on mammography (37%) than on ultrasound (76%). These findings quantify how imaging modality, lesion conspicuity, and dataset scale materially affect both segmentation and diagnostic performance, and they underscore the necessity of transparent, modality-aware reporting — including recall/sensitivity for the malignant class, not just aggregate accuracy — when evaluating deep learning systems intended to support breast cancer screening.