A deep learning-based hybrid approach for breast cancer detection

Breast cancer is the most commonly diagnosed type of cancer among women, accounting for approximately 30% of all cancer cases. Accurate identification of whether tumors in breast tissue are benign or malignant is of vital importance. Mammographic imaging is the most frequently used method for such diagnoses. However, the interpretation of mammograms can be time-consuming and may lead to erroneous results due to subjectivity in the evaluation process. It is well-established that early and accurate diagnosis of breast cancer significantly enhances treatment success and reduces mortality rates. Consequently, there has been a growing interest in artificial intelligence-based approaches to accelerate diagnostic processes and improve accuracy. In this study, deep learning models are employed to detect complex patterns and anomalies in mammographic images. A hybrid model based on Convolutional Neural Network (CNN) and Feedforward Neural Network (FNN) is proposed, aiming to achieve both high accuracy and rapid detection in breast cancer diagnosis. A total of 13,128 mammogram images (5,970 benign and 7,158 malignant) from the Digital Database for Screening Mammography (DDSM) were utilized. To prevent overfitting during the training phase and enhance data diversity, data augmentation techniques—including random rotation, translation, and scaling—were applied. These transformations generated dynamically altered examples for each mini-batch during training. The augmented images were then classified using the proposed CNN+FNN-based hybrid model. To ensure the reliability of classification accuracy, different train-test split ratios (50%-50%, 60%-40%, 70%-30%, and 80%-20%) were tested. Furthermore, k-fold cross-validation (k=5) was employed to comprehensively assess the model's performance. The proposed method achieved outstanding classification performance, including 99.75% accuracy, 99.77% precision, 99.63% recall (sensitivity), 99.77% F1-score, 99.89% specificity, and 99.49% Matthews Correlation Coefficient (MCC).