Researchers from Korea developed a deep learning model that can differentiate between multiple sclerosis (MS) and neuromyelitis optica spectrum disorder (NMOSD).
”This compact model may aid in the differential diagnosis of MS and NMOSD in clinical practice,” they wrote in a report that they published in the journal Scientific Reports.
Read more about the differential diagnosis of NMOSD
Continue Reading
The team developed the model using brain magnetic resonance imaging (MRI) data.
The model was based on a modified ResNet18 convolution neural network, which was trained using 5-channel images. These were obtained by selecting 5, 2-dimensional slices of 3-dimensional fluid-attenuated inversion-recovery (FLAIR) images.
The researchers used a total of 308 MRI scans from 86 patients with MS and 70 patients with NMOSD. Most patients with NMOSD were positive for the anti-aquaporin-4 antibody, and none of the patients were positive for the antimyelin oligodendrocyte glycoprotein antibody. The median Expanded Disability Status Scale of patients with NMOSD was 2.5 while that of patients with MS was 1.
The authors reported that the accuracy of the model was 76.1%. Its sensitivity was 77.3% and its specificity was 74.8%. Its positive predictive value was 76.9% while its negative predictive value was 78.6%. The area under the curve was 0.85.
When they applied a gradient-weighted class activation map to the model, the researchers saw that white matter lesions were the major classifier.
NMOSD is a rare chronic inflammatory disease affecting the brain and spinal cord. Although it shares many features with MS, it is a separate entity. The differential diagnosis between the 2 diseases is of great importance to ensure patients receive the right treatment as early as possible.
Reference
Seok JM, Cho W, Chung YH, et al. Differentiation between multiple sclerosis and neuromyelitis optica spectrum disorder using a deep learning model. Sci Rep. Published online July 19, 2023. doi:10.1038/s41598-023-38271-x
