Deep learning image analysis for breast cancer diagnosis

A medical company is trialling an AI system that can analyse mammograms for signs of breast cancer

Spotted: Kheiron Medical is set to launch trials of an artificial intelligence system, dubbed Mia, designed to aid in breast cancer diagnosis.

Many countries, including the UK, mandate that every mammogram should be read by two radiologists. However, a shortage of trained staff can make this standard difficult to achieve. Mia is a deep learning image analysis tool designed to act as an independent second reader.  

Mia was initially trained on around 500,000 scans from hospitals in Hungary. During training, Mia beat the average performance of a human radiologist. The system is now being tested with historic patient records to further assess and improve performance. Following analysis of historic data, the company’s next step will be to follow up with tests using current patients.

Healthcare is a major potential market for AI, but there are significant privacy obstacles that will need to be overcome. Kheiran hopes to alleviate privacy concerns through its extensive testing and partnerships with EU and UK health agencies. Kheiron is working with Britain’s NHS on further testing. They have also received funding from the EU Horizon 2020 research and innovation programme to roll out pilot sites in Europe.

Springwise has covered a number of innovations using AI to analyse medical data. Innovations covered recently include an AI-powered battlefield medical kit, a platform that can predict hospital admissions and an AI assistant that can detect heart attacks from a phone call.