http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
컨볼루션신경망을 이용한 시계그리기검사의 알츠하이머 치매 선별 가능성 확인
이시안 ( Lee Si-an ),홍준화 ( Hong Junhwa ),김나연 ( Kim Nayeon ),민혜민 ( Min Hyemin ),양하민 ( Yang Hamin ),이시현 ( Lee Sihyeon ),최서진 ( Choi Seojin ),박진혁 ( Park Jin-hyuck ) 대한인지재활학회 2023 대한인지재활학회지 Vol.12 No.2
Objective: This study was to examine the feasibility of the Clock Drawing Test (CDT) with convolutional neural networks (CNNs) for detecting Alzheimer’s disease (AD). Methods: 40 healthy older adults and 20 patients with mild AD conducted the CDT and then a total of 600 result images were established using augmentation techniques. 600 images were randomly allocated into training or test data sets, and 5-fold cross validation was applied. The CNN model was validated by 10 repetition tests, and its accuracy, sensitivity, and specificity were calculated. Results: No significant difference in demographic characteristics between both groups with exception of the Cognitive Impairment Screening Test (CIST) score. The CNN model achieved the accuracy of 85.0%, sensitivity of 87.5%, and specificity of 80.0%, whereas the CIST showed the accuracy of 80.0%, sensitivity of 87.0%, and specificity of 65.0%. Conclusion: Despite being based on a small amount of data, the CDT with CNNs showed higher accuracy than the CIST. Notably, it achieved a high specificity, which suggests that it has an advantage as a screening test in reducing false positives when disseminated in community settings and it could be a surrogate of the CIST.
이마앞엽 활성도를 이용한 인지 과제 난이도 예측 알고리즘 구축
전유진 ( Jeun Yu-jin ),김은진 ( Kim Eun-jin ),박도희 ( Park Do-hee ),이시안 ( Lee Si-an ),정다빈(1) ( Jung Da-bin(1) ),정다빈(2) ( Jung Da-bin(2) ),박진혁 ( Park Jin-hyuck ) 대한인지재활학회 2021 대한인지재활학회지 Vol.10 No.2
Objective: The purpose of this study was to provide training with a customized difficulty and confirm its effectiveness by using the difficulty used during training as the activity of prefrontal cortex using fNIRS. Methods: This study was conducted in two different stages. First, a study of the development of a difficulty prediction algorithm based on the activity of prefrontal cortex, and 45 subjects were recruited and data was collected. During the four cognitive tasks, the brain activity of prefrontal cortex was measured to create a dataset, and variables were selected thorough machine learning, algorithms were developed, and prediction accuracy was verified. Second, a study is to confirm the effectiveness of cognitive training by using algorithm, and one subject participated. Training was conducted five times for 30 minutes as a spatial cognitive task and before each training, the difficulty level to be used for training was measured using fNIRS and applied. Trial making test(TMT) by using fINRS was also evaluated at pre- and post-test. Results: In order to develop an algorithm, the variable was selected as the brain activity of VLPFC and cluster of two difficulty level group. The highest accuracy was 86%, which was logistic regression algorithm. As a result of applying the developed algorithm to the subject’s customized difficulty level for spatial cognitive training, activation was increased in prefrontal cortex and performance also was increased during training. Conclusion: These results suggested that the spatial cognitive training with subject-customized difficulty level using activity of brain would be useful in increasing in activation in prefrontal cortex and improving spatial cognition, which might have a positive effect on improving in brain efficiency.