Advanced Hybrid Techniques for Early Diagnosis of Mammogram Images: Combining Deep and Machine Learning Approaches

Digital image processing is a nascent field that emerged alongside computers. Its techniques vary by job. Applications that save, compress, distribute, and visualize images are different from those that analyze images for human visual perception. Digital image processing helps doctors diagnose patients using visual data in hospitals and clinics. In CAD systems, mammography-based early breast cancer detection is significant. CAD systems make unhealthy tissues more evident in medical images like digital chest X -rays and mammograms to facilitate early identification. Some clinicians prescribe digital mammography scans for women over 40 without a history of breast cancer. CAD systems employ low-dose X-rays that are mild on healthy breast tissue but conceal healthy and ill cells, making diagnosis harder. Thus, mammography second opinions are frequently needed to support and enhance medical decisions. Recent advances in deep and machine learning, particularly with computer-aided diagnostic tools, have improved diagnosis accuracy. Compared to prior methods, deep convolutional layers yield more resilient and recognizable features. Breast tumors are identified and classified using AlexNet and ResNet-18 models trained with AdaBoost. Mammograms are enhanced using an average filter first. Deep convolutional layers extract the most reliable deep characteristics. The AdaBoost technique classifies these traits, showing how deep learning and machine learning may improve diagnostic accuracy. The combination of ResnetNet-18 and Adaboost was the most effective hybrid model out of all the ones that were evaluated. The test results showed an accurateness of $\mathbf{8 1 . 5 \%}$, a compassion of $\mathbf{9 2 \%}$, and a specificity of $\mathbf{7 0 . 5 \%}$. The conclusions demonstration that using ResNet in conjunction with AdaBoost, a deep learning framework, improves diagnostic accuracy, sensitivity, and specificity.