Eye-Zheimer: A Deep Transfer Learning Approach of Dementia Detection and Classification from NeuroImaging
2020 IEEE 7th International Conference on Engineering Technologies and Applied Sciences (ICETAS)
(2020)
Helcy D. Alon
a
,
Michael Angelo D. Ligayo
b
,
Maribel A. Misola
c
,
Allan A. Sandoval
c
,
Marites V. Fontanilla
d
a Graduate Program, Technological Institute of the Philippines, Manila, Philippines
b Department of Electronics Engineering, Quezon City University, Quezon City, Philippines
c Department of Computer Engineering, FEU Institute of Technology, Manila, Philippines
d Department of Information Technology, St. Paul University, Quezon City, Philippines
Abstract: Dementia is a common term for memory loss, speech, problem-solving, and other cognitive skills that are serious enough to interfere with everyday life, and Alzheimer's is the leading cause of dementia. Alzheimer's disease is presumed to develop 20 years or more before symptoms occur, with degenerative changes that are unapparent to the person affected. The deep learning approach for early detection and Alzheimer's disease classification has recently gained significant attention. This study proposed disease detection trained by utilizing the YOLO v3 algorithm that aims to detect Alzheimer's disease based solely on Magnetic Resonance Imaging (MRI). Pascal VOC format and LabelImg tool are used for annotating the datasets, categorizing the image as non-demented and mild-demented. Model 4 was used in the system having 98.617% training accuracy, 98.8207% validation accuracy, and mAP of 96.17%. To test the accuracy of the used model, images of MRI scans are presented and it recorded 80% testing accuracy.