Breast cancer is one of the most common cancers diagnosed in women- and one of the most deadly. One in eight women will be diagnosed with breast cancer in their lifetime, with 85% of those diagnosed having no family history of breast cancer. In 2011, nearly 300,000 women were diagnosed with invasive and non-invasive breast cancer, 14% of which were expected to die from the cancer.
The stats, as staggering as they are, aren’t all bleak- death rates from breast cancer have been steadily decreasing since 1990 thanks to earlier detection from screenings. And a new screening tool developed by a 17 year old high-school junior shows promise for helping doctors diagnose more women earlier, so they can get the treatment they need to beat the cancer.
Brittany Wenger’s ‘Cloud 4 Cancer‘ app can detect early stages of breast cancer with 99.1% sensitivity, better than any existing commercial product. The app is a minimally invasive, accurate solution to diagnosing breast cancer–currently, the less invasive the procedure, the more prone to error, with accuracy of diagnosis improving with more invasive tools.
Wenger designed ‘Cloud 4 Cancer’ to work with findings from Fine Needle Aspirates (FNA), the least invasive way to determine if a mass is malignant or benign, using a computerized ‘brain,’ aka an artificial neural network:
Artificial neural networks detect patterns too complex to be recognized by humans and can be applied to breast mass malignancy classification when evaluating Fine Needle Aspirates (FNAs). This project teaches the cloud how to diagnose breast cancer by implementing a custom-crafted neural network that consumes FNA data collected by the University of Wisconsin to answer the question – is a mass malignant or benign?
Information regarding potential indicators of breast cancer is quantified in the dataset; specifically, clump thickness, single epithelial cell size, bare nuclei, mitoses, and five other attributes.
Just like a human brain that learns by repetition, the artificial neural network ‘learned’ to diagnosis breast cancer through repeated trials; Wagner tested 681 public FNA samples over 6,800 trials to determine the patterns across the nine cancer indicators that signal a malignant mass. Over the trials, the network found 222 instances malignant breast cancer with 93.67% accuracy and 417 instances of benign breast cancer with 99.52% accuracy. Only 2 samples that were actually malignant were categorized as benign, a false negative. Only 15 of the samples that were actually benign were categorized as malignant, and 42 total samples were ‘inconclusive.’
Personalizing the data, the 681 samples represent 681 women, of which only 42 would have to be tested again with a more invasive procedure. But because the neural network’s accuracy rate improves with the number of samples, this data, already with 99.1% sensitivity to correctly identifying malignant tumors, will be the ‘worst’ data set:
Neural networks learn just like humans. Therefore, at first the programs act like elementary school kids trying to diagnose breast cancer—they get almost all of the detections incorrect. As the neural networks run through training examples, they begin to detect patterns. By the end of training, the neural network is a seasoned physician and usually diagnoses over 90% of the tumors correctly.
Like the name suggests, Wagner has opened Cloud 4 Cancer to the cloud, hoping doctors around the world will enter their data sets– improving the accuracy of the network and more importantly, the accuracy of breast cancer diagnosis. Wagner hopes that neural networks could also be used to help diagnosis other diseases, like ovarian or prostate cancer.
Watch a video of Brittany explain Cloud 4 Cancer and visit her site here for more information:
PSFK has partnered with Boehringer Ingelheim to bring you a steady stream of inspiring news and ideas in the health and wellness space. Once each week, we will be posting an article on PSFK.com. If you would like to gain access to the full stream of content, please check out Boehringer Ingelheim’s Facebook page, where they are publishing a regular stream of inspiring and informative content.