Breast Cancer Detection: A Convolutional Neural Network based Approach for Robust Benign and Malignant Mass Identification Across Varied Breast Density
Ankita Patra, Santi Kumari Behera, J. Ramadevi, Prabira Kumar Sethy, Preesat Biswas, Himanshu Shekhar Bhoi · 2024
This paper presents a groundbreaking methodology for the identification of breast cancer using a Convolutional Neural Network (CNN) that has been specifically engineered to exhibit resilience in accurately distinguishing between benign and malignant masses in X-ray images. By effectively utilizing a meticulously curated dataset comprising 5040 images sourced from Kaggle, wherein each category is represented by an equal distribution of 2520 images, the Convolutional Neural Network (CNN) attains a commendable training accuracy of 84.13%. The remarkable ability of the model to discern intricate patterns among diverse breast densities is truly commendable, as it effectively tackles a crucial facet of breast cancer detection. During the testing process, it is noteworthy that the Convolutional Neural Network (CNN) consistently upholds a commendable level of accuracy, specifically measuring 75.92%. This outcome serves as a testament to the efficacy of CNN in real-world scenarios. This study highlights the inherent capacity of this convolutional neural network (CNN)-derived methodology as a potent instrument for the timely identification of breast cancer, presenting a promising pathway towards enhanced precision in diagnosis and, ultimately, better prognoses for patients.