Breast Cancer Disease Identification and Prediction based on Mammography Images Dataset: Comparative Analysis Approach

Abstract

Breast cancer remains one of the world’s deadliest health threats for women. Fortunately, early detection can mean the difference between life and. Doctors use mammography scans to detect tumours. But having 100 radiologists go through thousands of images every day is hard and may inaccurate so what we proposed here is A combination of K-Means clustering and Gaussian Mixture Models allows the system to concentrate on the facets of a mammogram that really matter by intelligently isolating the harmful areas. We passed them through ResNet50, a well-known model from deep learning, using transfer learning. Transfer learning allows us to use the existing model knowledge rather than training everything from scratch. We trained and evaluated the system on over 10,000 mammography images and the outcome was quite encouraging – The model attained an accuracy rate of 96.2% with strong precision and recall rates with confidence. This is much more useful for a real-world clinical application where doctors want to know how certain a prediction is. In conclusion, we believe that this framework can be truly beneficial to radiologists in their day-to-day work and can help reduce the diagnostic errors.