Russian and Indian Researchers Develop Low-Compute AI Algorithm for Breast Cancer Detection
The new system delivers high diagnostic accuracy while requiring only modest computing resources.
Researchers from Innopolis University, Saint Petersburg Electrotechnical University "LETI," and Jadavpur University have developed a new AI algorithm for breast cancer detection. The system analyzes digital microscope images of tissue samples using a neural network model that requires minimal computing power while achieving an accuracy rate of more than 99%.
The researchers achieved these results by combining two lightweight neural networks, SqueezeNet and ShuffleNet. This approach eliminates the need for the computationally intensive models commonly used in today's AI systems.
“Traditional computer-aided diagnostic systems based on transformers and deep ensembles are highly accurate, but they require enormous computing resources, making them difficult to deploy in healthcare facilities with limited infrastructure. Our approach addresses that challenge by combining two compact convolutional neural networks that work together,” said Dmitry Kaplun, lead programmer-mathematician at the AI Research Center of Innopolis University.
