Abstract #1372

Differentiation of Multiple Sclerosis and Neuromyelitis Optica Spectrum Disorders by Convolutional Neural Network

Akifumi Hagiwara^1,2, Yujiro Otsuka^1,3, Christina Andica¹, Shimpei Kato^1,4, Kazumasa Yokoyama⁵, Masaaki Hori^1,6, Shohei Fujita^1,4, Koji Kamagata¹, Ryusuke Irie^1,4, Saori Koshino^1,4, Tomoko Maekawa^1,4, Toshiaki Akashi¹, Akihiko Wada¹, Kanako Kunishima Kumamaaru¹, Takuya Haruyama^1,7, Syo Murata¹, Nobutaka Hattori⁵, and Shigeki Aoki¹

¹Radiology, Juntendo University School of Medicine, Tokyo, Japan, ²Radiology, UCLA, Los Angeles, CA, United States, ³Milliman Inc., Tokyo, Japan, ⁴Radiology, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan, ⁵Neurology, Juntendo University School of Medicine, Tokyo, Japan, ⁶Radiology, Toho University, Tokyo, Japan, ⁷Radiological Sciences, Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo, Japan

Multiple sclerosis (MS) and neuromyelitis optica spectrum disorders (NMOSD) are both neuroinflammatory diseases and have overlapping clinical manifestations. We developed a convolutional neural network that differentiates between MS and NMOSD based on multi-dynamic multi-echo sequence that measures R1 and R2 relaxation times and proton density. To avoid overfitting on a small dataset, we aimed to separate features of images into those specific to an image and those common to the group (i.e. MS or NMOSD) based on SqueezeNet. We used only common features for classification. Our model achieved a diagnostic accuracy of 80.7%.

This abstract and the presentation materials are available to members only; a login is required.

Join Here