Dear AuntMinnie Member,
A new deep-learning algorithm is showing promise for the detection of brain hemorrhage on noncontrast CT scans, according to research presented at last week's Society for Imaging Informatics in Medicine (SIIM) meeting in National Harbor, MD.
Researchers from the University of California, San Francisco (UCSF) presented their work on a convolutional neural network (CNN) algorithm designed to deliver high sensitivity for detecting small microhemorrhages in the brain with a low false-positive rate.
What's more, the algorithm has shown strong performance since it was put into clinical use in UCSF's emergency room -- demonstrating that real-world applications of deep learning can work. Find out how the group did it by clicking here, or visit our Artificial Intelligence Community at ai.auntminnie.com.
Breast pain and imaging overuse
Breast pain isn't necessarily a sign of cancer, but many women with the condition are referred to radiology for additional workup. A new study indicates this might not be a good idea.
Researchers from MD Anderson Cancer Center in Houston analyzed the outcomes of nearly 800 women who were referred for imaging workup for breast pain. They found that the referrals led to hundreds of imaging studies, even though the vast majority of women didn't have cancer. Annual screening mammography might be a better strategy, they believe.
Read more by clicking here, or visit the community at women.auntminnie.com.
New PET tracer for NET detection
Finally, visit our Molecular Imaging Community for an article on a new radiotracer for the detection of neuroendocrine tumors (NETs). A group from Switzerland and Germany tested a radiotracer based on gallium-68 that targets the somatostatin receptor subtype 2, which is commonly found on gastrointestinal and pancreatic tumors. Found out how well it worked by clicking here, or visit the community at molecular.auntminnie.com.
