Shea, Colin (NINDS)
Crainiceanu, Ciprian (Johns Hopkins University)
Sweeney, Elizabeth (Johns Hopkins University)
Shinohara, Russell (Johns Hopkins University)
Goldsmith, Arthur (Johns Hopkins University)
Software / Apps
This invention relates to methods and algorithms that incorporate information from multiple imaging modalities to identify, estimate the size, and track the time course of brain lesions. Subjects develop brain lesions over the natural course of a disease. Currently, lesions are measured and tracked by a trained neuroradiologist using slice-by-slice inspection, a slow process that is prone to human error and hard to generalize to large observational studies.
Researchers at the National Institute of Neurological Disorders and Stroke (NINDS) and Johns Hopkins University have created a novel method for automatically detecting brain lesion incidence in longitudinal brain magnetic resonance imaging (MRI) studies (data collected at multiple visits). The software embodying the invention, Subtraction Based Logistic Inference for Modeling and Estimation (SuBLIME), was applied on MRI studies with only T2-weighted images, and studies with T2-weighted, T1-weighted, fluid attenuated inversion recovery (FLAIR) and proton density (PD) MRI modalities.
The technology can be extended to additional diseases of the brain (including stroke, small vessel disease, tumors, metabolic disorders, and inflammatory disorders), include additional imaging modalities (such as CT, PET, and PET-CT). In addition, the algorithm could be extended to whole-brain segmentation and to account for shrinking and disappearing lesions. Other body parts where lesions need to be tracked could also be included.
Automated detection and characterization of brain lesions.