대한신장학회

Skip Navigation: Skip to contents

KOR
ENG
전체메뉴보기

간행물 검색

현재 페이지 경로

HOME
간행물
간행물 검색

논문분류	춘계학술대회 초록집
제목	Deep Neural Network for Estimating Glomerular Filtration Rate
저자	Mun Hee Chai, Byoung-Geun Han, Han Wul Shin, Jin Jae Jeong, Jae Seok Kim, Seong Ok Choi, Jun Young Lee, Jae Won Yang
출판정보	2019; 2019(1):
키워드	GFR \| MDRD \| DNN
초록	The glomerular filtration rate (GFR) is a typical method to estimate the patient’s renal function status. Several methods have been proposed, including the Modification of Diet in Renal Disease (MDRD) method and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation that is based on linear regression for estimating GFR. This study aimed to develop a deep neural network (DNN) to estimate GFR. 241 patients were enrolled from Wonju Severance Christian hospital. All subjects measured DTPA Renogram, serum creatinine level. We constructed the DNN model composing three hidden layers: first, second, and third hidden layers included 40, 20, and 10 nodes. We determined the activation function of DNN as hyperbolic tangent. By using delta GFR (Measured GFR-Estimated GFR), Root squared error (RSE), P30, and accuracy of classification CKD patients (GFR < 60 mL/s per 1.73 m2), the performance of DNN model was measured. Means of the median difference of MDRD, CKD-EPI, and DNN methods in five CV sets were 2.354 (– 0.75 to 6.92), 2.862 (– 1.12 to 7.98), and 1.866 (– 2.34 to 4.68), respectively. Averages of median RSE of MDRD, CKD-EPI, and DNN methods were 19.45, 18.9, and 16.78, respectively. In P30, MDRD showed the largest ratio among the three methods. In a classification of GFR < 60 mL/s per 1.73 m2, means of the accuracy of MDRD, CKD-EPI, and DNN methods were 83.0, 84.3, and 85.1, respectively By using DNN, we could proposed the most accurate estimated GFR in Korean patients.
원문(PDF)	PDF 원문보기

목록

위로가기