New Jersey residents may soon have a lower risk of having their pneumonia misdiagnosed. Imaging software that is powered by artificial intelligence developed by researchers at Stanford University may be used to help physicians more accurately diagnose pneumonia.
The software, CheXnet, is a neural network that can evaluate images based on certain parameters. The neural network was trained using 112,120 images of chest X-rays with descriptors for up to 14 different medical conditions, one of which was pneumonia. Following one month of training, the results of the software exceed those of the other computer-based procedures that had been used to identify pneumonia.
The software was also tested against four radiologists at the university. Tested factors included sensitivity and specificity. Sensitivity refers to the amount of correctly identified positives while specificity refers to the amount of accurately detected negatives.
Using the software is relatively simple. Physicians are able to input an X-ray of a patient’s lungs and receive a numerical likelihood of whether a pneumonia infection is in the lungs. A color map indicates the infection levels in the tissue. With the information provided by the software, physicians are able to determine a course of treatment.
The developers of the software want their research to be applied all over the world. It is estimated that two-thirds of the world lack radiology diagnostic tools that can provide accurate results.
Individuals whose medical conditions are misdiagnosed may have legal recourse. A personal injury attorney who practices medical malpractice law may pursue financial damages against negligent doctors. Financial settlements could cover pain and suffering as well as any unnecessary medical expenses that resulted from a misdiagnosis.