Brain MRI has been established as the main diagnostic modality for assessing a variety of neurological disorders, and hundreds of these studies may be performed on a daily basis in a busy clinical practice. As a result, there is significant interest in automated image processing and classification, according to Dr. Kambiz Nael from the Icahn School of Medicine at Mount Sinai in New York City.
In a joint partnership between the Icahn School of Medicine and Siemens Healthineers, the researchers developed an AI 3D pipeline that includes landmark detection for brain positioning and coverage, deep learning-based skull stripping, and a set of sequence-independent detection networks to classify clinical brain MRI scans as normal or abnormal, Nael said.
"Our proposed intelligent pipeline accurately identifies abnormal brain MRIs from multiple contrasts of individual patients," he told AuntMinnie.com. "If its potential is realized, it can be used as a clinical tool to flag abnormal MRIs, allowing for improved triage and timely interpretation of abnormal scans in a busy and large clinical practice."
The researchers anticipate that the algorithm will yield improvements to their institution's clinical workflow, including flagging of abnormal brain MRI scans in less than two seconds from study completion.
"MRI technologists can see the red flags and can call the neuroradiologists to accelerate the interpretation and potentially avoid sending patients with acute neurological findings (such as stroke) home," he said.
Furthermore, all brain MRIs performed throughout their institution could be stratified and classified, enabling neuroradiologists to prioritize reading of abnormal scans and avoid delays in interpreting these studies at the beginning of a day or after weekends and holidays, according to Nael.
Stop by this talk to learn more.
