New ai model detects breast cancer with near-perfect accuracy
Early detection of breast cancer can save lives — and a new artificial intelligence system is setting a new benchmark for accuracy. researchers have developed an optimized multi-instance gates-controlled deep unfolding network, named MGCDUN–WOA, that identifies breast cancer from mammograms with exceptional precision.
Tested on two major public datasets, Inbreast and DDSM, the model achieved a 99.9% accuracy and 99.8% recall rate, outperforming existing state-of-the-art methods.
How Technology work
The process begins with hybrid fast conventional bilateral filtering, which cleans mammogram images by removing noise and artifacts. a cascaded graph convolutional network then performs precise segmentation of potential cancer regions. next, the multi-instance gates-controlled deep unfolding network (MGCDUN) extracts key diagnostic features using a multi-head transformation (MHT) technique.
To further enhance performance, the researchers used the walruses optimization algorithm (WOA), a bio-inspired method that fine-tunes model parameters for maximum efficiency and robustness. The result is a system that not only detects tumors with near-perfect accuracy but also operates faster and more reliably than traditional machine learning techniques.
Experts say this model could become a powerful tool for radiologists, reducing diagnostic errors, improving early-stage detection, and leading to timelier treatment and better patient outcomes.The study highlights how deep learning, when coupled with advanced optimization algorithms, continues to reshape modern healthcare diagnostics.
Source: https://www.kaggle.com/datasets/ramanathansp20/inbreast-dataset
