Objective: To explore the brain-predicted age difference (Brain-PAD) in patients with type 2 diabetes mellitus (T2DM) by a machine learning prediction model based on structural magnetic resonance ( sMRI) in the Southwest University Adult Lifespan Dataset (SALD) , and to reveal the relationship between Brain-PAD and dura- tion of T2DM and cognition . Methods: Group comparisons about demographic variables and cognitive function were conducted respectively in local database of 104 T2DM patients and 83 healthy controls (HC) . The prediction model via Gaussian process regression (GPR) was constructed by training sMRI data of 329 healthy volunteers in SALD , then its performance was validated and evaluated . Furthermore , Brain-PAD ( predicted age-chronological age) in the local database was calculated . Group comparisons of Brain-PAD between T2DM patients and HCs were conducted by Mann-Whitney U test. Finally , Pearson correlation coefficient (r) was calculated between Brain-PAD and duration of disease and cognition . Results: Poor performance in auditory verbal learning test (AVLT)-delayed recall , AVLT-recognition , symbol digital modalities test (SDMT) (P < 0. 05) , and increased Brain-PAD were ob- served in T2DM patients , compared with HCs [1 . 619 ( - 4. 001 , 8. 272) years vs - 1 . 289 ( - 4. 128 , 4. 134) years , Z = 2. 056 , P = 0. 034] . Notably , the median of Brain-PAD in T2DM group was positive , indicating that the brain of T2DM patient maybe relatively “older”than his chronological age . Brain-PAD in T2DM group was as- sociated with performance in AVLT-immediate recall ( r = 0. 291 , P = 0. 003) , AVLT-delayed recall ( r = 0. 248 , P = 0. 011) , SDMT( r = 0. 376 , P = 0. 001) and trail making test (TMT)-A ( r = - 0. 206 , P = 0. 036) . However , the relationships between Brain-PAD and duration of T2DM were not explored . Conclusion Decreased cognitive function in patients with T2DM is demonstrated in this study . The machine learning prediction model based on sMRI supports the identification of brain aging objectively in patients with T2DM .

Objective To determine the relationship between serum soluble Klotho (sKL) level and adverse outcome in maintenance hemodialysis (MHD) patients.Methods One hundred and twenty nine cases of MHD patients were collected prospectively.Serum sKL was detected by ELISA.Abdomen lateral plain was used as a criterion to determine the abdominal aortic calcification.The abdominal aortic calcification score (AAC) was calculated.Cox regression analysis was used to determine the risk factor of cardiovascular death (CVD) in MHD patients.Kaplan-Meier showed the relationship between sKL and CVD in MHD patients.Results There were 27 cases (20.9％) of allcause death and 19 cases (14.7％) of cardiovascular death.The median sKL was 612.6(379.2-816.6) nig/L,and log[iPTH] was an independent factor of sKL concentration.Low sKL had high AAC and CVD death rate.Kaplan-Meier method showed that the all-cause death rate was similar between two groups,and CVD death rate increased significantly in low sKL patients (P=0.036).Cox regression indicated that lower sKL level was associated with high CVD death rate [OR=0.352,95％CI(0.127-0.977),P=0.045].After adjustment for the general condition,biochemical indicators,the relationship still existed [OR=0.331,95％ CI (0.117-0.933),P=0.037].In no or mild vascular calcification patients (AAC ≤4),compared with high sKL patients,low sKL patients had no significant difference rate in all-cause mortality.The CVD mortality was significantly higher in high sKL (P=0.035) compared with low sKL.In severe calcification group (AAC ＞ 4),all-cause death and CVD death rates were similar between different sKL groups (P=0.991 and 0.522,respectively).Conclusions Lower sKL has the high CVD death rate and sKL level decreasing is an independent risk factor for CVD death in MHD patients.The lower sKL concentration in MHD patients with no or mild calcification may predict CVD mortality.This study suggests that sKL levels may be helpful in predicting the outcome of patients with MHD.

Objective To explore the association of fibroblast growth factor-23 (FGF23) with abdominal aortic calcification(AAC) and adverse outcomes in maintenance hemodialysis patients.Methods One hundred and fourteen cases of MHD patients were collected prospectively.Serum intact FGF23 was detected by ELISA.Abdomen lateral plain was used as a criteria to determine the abdominal aortic calcification and the abdominal aortic calcification score was counted.Logistic regression analysis was used to determine the risk factors of AAC.Kaplan-Meier analysis was applied to compare the survival rate among different groups and COX regression analysis was used to determine the association of FGF23 and mortality in MHD patients.Results Seventy-six patients present abdominal aortic calcification.The median of AACS was 4.0(0.0,11.0).The median level of FGF23 was 7277.4(2535.0,9990.8) pg/ml.The median follow-up duration was 72.0(67.8,72.8) months.During the follow-up,22 patients (19.3％) died of all-cause death and 17 cases (14.9％) died of cardiovascular diseases.Serum FGF23 level was positively correlated with AAC (r=0.285,P=0.002).Logistic regression analysis showed that longer age (OR=1.059,95％CI:1.020-1.100,P=0.003) and dialysis vintage (OR=I.009,95％CI 1.000-1.017,P=0.039),smoking history (OR=3.010,95％CI 1.177-7.696,P=0.021) and higher FGF23 level(OR=2.831,95％CI 1.010-7.937,P=0.048) were independent risk factors of moderate to severe AAC in MHD patients.Kaplan-Meier survival curves showed that the patients with AACS≥ 5 had significantly higher all-cause mortality(P=0.028) and CVD mortality (P=0.035) than those with AACS ＜ 5.However,the Kaplan-Meier analysis showed no significant difference regarding the level of serum FGF23 with the all-cause and CVD mortality.Cox regression demonstrated that FGF23 was not associated with increased mortality risk,neither in crude nor in multivariate adjusted models.Conclusions Abdominal aortic calcification had a high prevalence in MHD patients.The all-cause and CVD mortality was higher in patients with moderate to severe AAC.FGF23 was an independent risk factor of moderate to severe AAC,but it can't yet be a predictor for the allcause and CVD mortality of MHD patients.

Objective To investigate the relationship between serum phosphorus variability and mortality in maintenance hemodialysis (MHD) patients. Methods A total of 502 MHD cases from Renji hospital hemodialysis center were registered in Shanghai Registry Network from January 2007 to April 2015. They were recruited with general information, laboratory results and outcomes. According to their median of coefficient of variation (CV) of blood phosphorus, the patients were divided into high variation group (CV≥0.226 mmol/L) and low variation group (CV<0.226 mmol/L). The relationship of serum phosphorus CV with all?cause mortality and cardiovascular disease mortality was assessed respectively. Results The average age was (63.9±14.6) years, the median dialysis age was 82.0 (43.0, 139.0) months, 118 patients (23.5%) died for all cause and 64 patients (12.7%) died for cardiovascular disease. Compared with patients in low phosphorus variation group, patients had a higher all?cause mortality in high phosphorus variation group (27.7% vs 19.3%, P=0.028). Higher cardiovascular disease mortality was observed in high variation group as well, but this difference was no statistical significant (15.4% vs 10.0%, P=0.082). COX regression analysis showed that >60 years of age (HR=2.762, 95%CI 1.707?4.468, P<0.001), low hemoglobin (HR=0.466, 95%CI 0.317?0.686, P<0.001), low albumin (HR=0.555, 95%CI 0.366?0.840, P=0.005), high CV of phosphorus (HR=1.479, 95%CI 1.023 ? 2.139, P=0.037) were independent risk factors for all ? cause mortality. Moreover, >60 years of age (HR=2.666, 95%CI 1.469?4.837, P=0.001), low hemoglobin (HR=0.480, 95%CI 0.238?0.801, P=0.005), and high CV of phosphorus (HR=1.655, 95%CI 1.003?2.729, P=0.049) were independent risk factors for cardiovascular disease mortality. There was no significant statistical difference between patients phosphorus on target and patients phosphorus below target in all?cause disease mortality (P=0.065) and cardiovascular disease mortality (P=0.425). High variation group whose phosphorus on target had higher all?cause mortality and cardiovascular disease mortality than those in low variation group (29.2% vs 16.9%, P=0.047; 15.0% vs 6.0%, P=0.033). Kaplan?Meier method showed that patients with high phosphorus variation had higher all?cause (P=0.023) and cardiovascular disease mortality (P=0.047) than patients with low phosphorus variation. Conclusions The high CV of phosphorus is independently correlated with all?cause and cardiovascular disease mortality. Patients with standard ? reaching phosphorus in the low variation group have a lower mortality. A serum phosphorus level sustainably reaching the standard may improve the survival in MHD patients.

Semisupervisedmakesfulluseoflargeamountsofunlabeledsamplestomakeuptheinsufficiency of labeled samples. Since it is difficult to obtain a large number of accurate labeled samples and it is a good way for modeling by using a small amount of labeled samples or a large number of inaccurate samples, we proposed a new method named as semi-supervised partial least squares ( SS-PLS) to optimize model based on semi supervised learning. We used 211 samples of tobacco near infrared spectrum and sensory evaluation for modeling and used SS-PLS method to optimize tobacco sensory evaluation model. In the optimized model, the coefficient of determination ( R2 ) can reach up to 90%, the ratio of performance to deviation ( RPD) can reach up to 3 . 0 , and the standard error of cross validation and the standard error of prediction ( SECV and SEP) are below 1. 0. We divided the original sensory evaluation and SS-PLS optimized data into three grades of excellent, medium and poor in accordance with the fixed threshold, the result using projection model of based on principal component and Fisher criterion ( PPF ) shows that the classification of SS-PLS optimized data is better than the original sensory evaluation data. The SS-PLS method can solve the data representation problem of using small sample set for modeling and provides a new chemometrics method for near infrared spectroscopy modeling in case of obtaining a large number of accurately labeled samples is difficult.