Computer-aided diagnosis of head and neck lesions from non-Gaussian diffusion MRI signal patterns
Mami Iima 1 , Akira Yamamoto 1 , Denis Le Bihan 2,3 , Shigeru Hirano 4 , Ichiro Tateya 4 , Morimasa Kitamura 4 , and Kaori Togashi 1
Department of Diagnostic Imaging and Nuclear
Medicine, Graduate School of Medicine, Kyoto University,
Kyoto, Kyoto, Japan,
Brain Research Center, Graduate School of Medicine,
Kyoto University, Kyoto, Kyoto, Japan,
CEA-Saclay Center, Gif-sur-Yvette Cedex, France,
of Otolaryngology, Head and Neck Surgery, Graduate
School of Medicine, Kyoto University, Kyoto, Kyoto,
This prospective study included 46 patients suspected of
head and neck tumors. They were scanned using a RS-EPI
diffusion MRI sequence implemented on a 3T MRI scanner.
Images were analyzed with a new approach algorithm which
enables automatic classification of tumor types from a
"signature index" (S-index) directly based on the
non-Gaussian diffusion signal pattern obtained from 2
gkey b valuesh. This computer-assisted diagnostic
algorithm allowed malignant and benign lesions to be
differentiated with a high AUC (0.89). The lesion
S-index histogram and 3D display also highlighted the
importance of tumor heterogeneity.
This abstract and the presentation materials are available to members only;
a login is required.