Software controlled by artificial intelligence may soon help doctors in Arizona and across the U.S. diagnose pneumonia more accurately. The software could also improve health care treatment in parts of the world that have limited access to radiologists.
CheXnet, which was created by researchers at Stanford University, uses a convolutional neural network to process chest images according to a specific set of parameters. Scientists trained the software's AI brain by having it process 112,120 chest X-ray images labeled with up to 14 separate diseases, including pneumonia. After four weeks, CheXnet was outperforming older types of diagnostic software used to help diagnose pneumonia. Researchers then tested the software against four skilled radiologists and found that it was more accurate than humans at spotting the disease.
According to the researchers, CheXnet works by having doctors scan a chest image into the software. The program analyzes the image and issues a percentage probability that the patient has pneumonia and a color map of infectious areas in the lungs. Doctors can then use that data to help with their diagnosis and treatment plan. Medical experts say that CheXnet could significantly reduce the number of pneumonia misdiagnoses throughout the world, which could save lives.
Pneumonia misdiagnosis can lead to a worsened medical condition and even death. Patients who have suffered harm due to a doctor's failure to diagnose this or other diseases may find it beneficial to contact an attorney who has experience with medical malpractice cases in order to see how best to seek compensation for the additional medical expenses and other losses that have been incurred.
Source: Fast Company, "New AI Can Diagnose Pneumonia Better Than Doctors", Jesus Diaz, Nov. 27, 2017