Scientists from Innopolis University, St. Petersburg Electrotechnical University "LETI," and Jadavpur University in India have created an AI model for breast cancer diagnosis with 99.72% accuracy
To create the new diagnostic system, the scientists combined two compact neural networks — SqueezeNet and ShuffleNet. This approach allowed them to maintain high accuracy in breast cancer detection without the use of resource-intensive models that require powerful hardware.
During testing on international histopathology image datasets, the algorithm demonstrated accuracy of up to 99.72%. It distinguishes between healthy and diseased tissue and identifies different stages of the disease.
The researchers also confirmed that the results are not random: the algorithm analyzes precisely those image areas that are important for diagnosis.
