Photo: Google Science Fair
An artificial “brain” web-based application built by a 17 year old from Florida can assess tissue samples for signs of breast cancer with amazing accuracy.Brittany Wenger built a cloud-based neural network — a computer program coded to process data and detect patterns — that is currently 99.1 per cent sensitive to malignancy.
“This is really important because currently the least invasive form of biopsy is actually the least conclusive, so a lot of doctors can’t use them,” Wenger, who began building neural networks in seventh grade, told NBC News.
Artificial neural networks “learn” by analysing data sets and perform better as they process more data. Eventually they’re able to detect patterns in data that are too complex for human brains or other types of programs to analyse.
Wenger aimed to improve the accuracy of the least invasive diagnostic procedure, called fine needle aspirate, so women don’t have to undergo a second biopsy with a bigger needle or even surgery.
Wenger’s network, programmed using Java, provided better results than three commercially available networks as it gave correct diagnoses for 94 per cent of the cases and correctly identified more than 99 per cent of the cancerous cases.
Less than one per cent of the answers were false negatives that gave diagnoses for cancerous lumps when the growths were actually benign. Commercial neural networks, on the other hand, had a false negative rate of about 5 per cent.
Her app, called Cloud4Cancer, is available online so that doctors can enter their own data. She also thinks her approach may be used to make neural networks that diagnose other cancers.
Her project won grand prize at the 2012 Google Science Fair, earning her an internship at one of the fair’s sponsors, a trip to the Galapagos Islands, a trophy made of white Lego bricks and a $50,000 scholarship for college.
Business Insider Emails & Alerts
Site highlights each day to your inbox.