Abstract:
Introduction: Breast cancer is a type of cancer that affects the fibroglandular cells in the breast. According to a 2018 estimate, the number of new cases of breast cancer worldwide is expected to rise to 29.5 million by 2040. In Rwanda, breast cancer is the most common cancer among women, and delays in diagnosis and treatment are believed to contribute to the high number of cases. Mammography screening is typically performed on women of a certain age to detect early signs of cancer. However, this method places a heavy burden on radiologists who are already scarce in number. Moreover, each person's ability to detect lesions in mammography images varies, and if a radiologist fails to detect the cancerous growth, additional exams may prove to be expensive in terms of both finances and lives. Aim. This study aimed to develop a deep-learning AI program that classifies medical images as normal or abnormal, aiding radiologists in their diagnoses. Methodology. Our project utilized Python's built-in methods and packages to create, train, and evaluate a deep-learning model. The model was trained on 1200 images from King Faisal Hospital, all of which were captured in 2022. The Residual Neural Network 50 was the focal point of our approach. Results. Our results showed a sensitivity probability of 1.00, an accuracy of 0.78, a precision of 0.69, and an F1 score of 0.52. Conclusion. The model is currently performing well. However, it is highly recommended to continue the project by comparing it to other deep learning models or increasing the dataset size to achieve optimal performance. This is especially important in a medical setting, where even a small error can result in the loss of life. Therefore, it's crucial to ensure that the model is thoroughly tested and validated before deploying it.
