Deep learning based low-cost high-accuracy diagnostic framework for dementia using comprehensive neuropsychological assessment profiles

• DOI: 10.1186/s12877-018-0915-z
• Publication Date: 2018-10-03T14:02:29Z
• Authors: Hyun-Soo Choi; Jin Yeong Choe; Hanjoo Kim; Ji Won Han; Yeon Kyung Chi; Kayoung Kim; Jongwoo Hong; Taehyun Kim; Tae Hui Kim; Sungroh Yoon; Ki Woong Kim
• Keywords: Robustness; Neuropsychological Assessment

The conventional scores of the neuropsychological batteries are not fully optimized for diagnosing dementia despite their variety and abundance information. To achieve low-cost high-accuracy diagnose performance using a battery, novel framework is proposed response profiles 2666 cognitively normal elderly individuals 435 patients who have participated in Korean Longitudinal Study on Cognitive Aging Dementia (KLOSCAD). key idea to propose cost-effective precise two-stage classification procedure that employed Mini Mental Status Examination (MMSE) as screening test KLOSCAD Neuropsychological Assessment Battery diagnostic deep learning. In addition, an evaluation redundant variables introduced prevent degradation. A missing data imputation method also presented increase robustness by recovering information loss. neural networks (DNNs) architecture validated through rigorous comparison with various classifiers. k-nearest-neighbor has been induced according framework, DNNs two stage show best accuracy compared other Also, 49 were removed, which improved suggested potential simplifying assessment. Using this we could get 8.06% higher than MMSE alone 64.13% less cost KLOSCAD-N alone. be applied general early detection programs improve robustness, preciseness, cost-effectiveness.