Scientists at Stanford University have developed a new artificial intelligence-powered imaging software designed to help doctors diagnose pneumonia with greater accuracy than ever before. It’s a great example of how AI is helping humans do their jobs better and faster. Its name is CheXnet and it is a convolutional neural network–a type of neural network that is designed to process images according to a limited set of parameters. In the case of CheXnet, the research team led by Stanford adjunct professor Andrew Ng, started by training the neural network with 112,120 chest X-ray images that were previously manually labeled with up to 14 different diseases. One of them was pneumonia. After training it for a month, the software beat previous computer-based methods to detect this type of infection. The Stanford Machine Learning Group team pitted its software against four Stanford radiologists, giving each of them 420 X-ray images. This graphic shows how the radiologists–represented by the orange Xs–did compared to the program–represented by the blue curve. ChexNet is tested against radiologists on sensitivity (which measures the proportion of positives that are correctly identified as such) and specificity (which measures the proportion of negatives that are correctly identified as such). Importantly, it has a simple user experience: Doctors input an X-ray of lungs, and automatically get the numeric probability of those lungs being infected with pneumonia or not. They also get a color map that highlights the level of infection throughout the tissue. Using these insights, the doctors can then make decisions about how to approach treatment. The team hopes to see its research applied not only in hospitals but all around the world; an estimated two-thirds of the global population doesn’t have access to accurate radiology diagnostic tools. The researchers “hope that this technology can improve healthcare delivery and increase access to medical imaging expertise in parts of the world where access to skilled radiologists is limited,” the announcement reads. As IEEE Spectrum’s Tekla S. Perry writes, the accuracy of the CheXnet program can make a tremendous difference in emergency rooms around the world–in fact, “failure to promptly recognize and treat bacterial pneumonia may lead to significant morbidity and mortality.” She describes how her 18-year-old son went to Stanford Medical Center’s ER with “extremely high fever and cough” twice but the doctors said the X-rays didn’t show any sign of pneumonia. It was only after a routine reevaluation of X-ray images when other doctors realized that he could have pneumonia, which indeed was the case. And misdiagnosis can go both ways: In an October 2010 paper, researchers discovered that “72% of patients were misdiagnosed with pneumonia upon readmission to the same hospital.” I, for one, agree with Perry: Next time I go to a hospital, I want AI fighting on my side. Source
