Background: Predicting hypoglycemia while maintaining a low false alarm rate is a challenge for the wide adoption of continuous glucose monitoring (CGM) devices in diabetes management. One small study suggested that a deep learning model based on the long short-term memory (LSTM) network had better performance in hypoglycemia prediction than traditional machine learning algorithms in European patients with type 1 diabetes. However, given that many well-recognized deep learning models perform poorly outside the training setting, it remains unclear whether the LSTM model could be generalized to different populations or patients with other diabetes subtypes. Objective: The aim of this study was to validate LSTM hypoglycemia prediction models in more diverse populations and across a wide spectrum of patients with different subtypes of diabetes. Methods: We assembled two large data sets of patients with type 1 and type 2 diabetes. The primary data set including CGM data from 192 Chinese patients with diabetes was used to develop the LSTM, support vector machine (SVM), and random forest (RF) models for hypoglycemia prediction with a prediction horizon of 30 minutes. Hypoglycemia was categorized into mild (glucose=54-70 mg/dL) and severe (glucose<54 mg/dL) levels. The validation data set of 427 patients of European-American ancestry in the United States was used to validate the models and examine their generalizations. The predictive performance of the models was evaluated according to the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: For the difficult-to-predict mild hypoglycemia events, the LSTM model consistently achieved AUC values greater than 97% in the primary data set, with a less than 3% AUC reduction in the validation data set, indicating that the model was robust and generalizable across populations. AUC values above 93% were also achieved when the LSTM model was applied to both type 1 and type 2 diabetes in the validation data set, further strengthening the generalizability of the model. Under different satisfactory levels of sensitivity for mild and severe hypoglycemia prediction, the LSTM model achieved higher specificity than the SVM and RF models, thereby reducing false alarms. Conclusions: Our results demonstrate that the LSTM model is robust for hypoglycemia prediction and is generalizable across populations or diabetes subtypes. Given its additional advantage of false-alarm reduction, the LSTM model is a strong candidate to be widely implemented in future CGM devices for hypoglycemia prediction.

Introduction: Simple non-invasive biomarker is urgently needed to detect the largely silent osteopenia in order to prevent osteoporosis-related fracture later in life. The accumulation of advanced glycation end products (AGEs) has been related to reduced bone density and osteoporotic fractures. Whether lens autofluorescence (LAF) based AGEs (LAF-AGEs) measurement could be used to assess the risk of osteopenia is aimed to investigate in this paper. Methods: Through routine health examination, 368 individuals under the age of 50 were enrolled. A dual-energy X-ray absorptiometry (DXA) device was used to measure bone mineral density (BMD) of the forearm and determine osteopenia. AGE levels were derived with LAF along with the other demographic and laboratory parameters. After deriving the age-adjusted AGE levels (AALs), a linear regression analysis and an ordered logistic regression analysis were applied to examine the associations between osteopenia and LAF-AGEs as well as AALs. Results: Negative correlations (Pearson r = − 0.16, p < 0.001) were found between LAF-AGEs and T-scores. Higher AALs were significantly associated (p = 0.004) with escalated level of osteopenia in the ordered logistic analysis. Discussion: After reviewing the relevant studies, it is concluded that LAF-AGE is a more stable measure of long-term metabolic dysfunction than circulating AGE. LAF-AGEs are a valid, practical and non-invasive parameter for osteopenia risk evaluation. Further studies with longer follow-up will be helpful to clarify its effectiveness for osteoporosis risk assessment.

Objective: To compare the clinical effects of single-incision robot-assisted laparoscopic radical prostatectomy (RARP) with and without extraperitoneal special channel device. Methods: The clinical data of 70 patients who had undergone RARP in the Robotic Minimally Invasive Surgery Center of Sichuan Provincial People's Hospital from September 2020 to February 2021 were analyzed retrospectively, including 29 cases who were operated on without special channel device (group A) and 41 cases with special channel device (group B). All operations were performed by robot-assisted single-incision retrograde bladder neck exfoliation via extraperitoneal approach in patients by the same operator. The operation time, intraoperative blood loss, the bladder neck urethral anastomosis time, postoperative hospital stay, postoperative exhaust time, positive rate of incisal margin, indwelling time of urinary catheter, retention rate of postoperative erectile function, satisfaction rate of immediate postoperative urine control, positive rate of postoperative lymph node pathology, incision length, treatment cost and the rate of prostate specific antigen (PSA)lower than 0.2 μg/L at 6 weeks after operation were compared between the two groups. Results: All 70 cases were operated successfully. The difference of age[ (68.9±3.9) vs (69.4±5.4) years], preoperative PSA level[14.1(6.3, 19.8)vs13.7(5.8, 18.1)μg/L], prostate volume[44.8(30.7,172.6)vs 56.3(40.9,163.4)ml ] of the two groups was not statistically significant(all P>0.05). The difference of operation time [ (59.1±18.5) vs (59.6±18.0) min ], intraoperative blood loss [93(66,198)vs 95(68,203) ml ], bladder neck urethral anastomosis time [ (12.6±1.3) vs (13.7±2.8) min ], postoperative hospital stay [ (8.1±2.3) vs (9.1±1.3) d], postoperative exhaust time [ (1.4±0.6) vs (1.3±0.6) d], positive rate of incisal margin (20.7% vs 19.5%), indwelling time of the urinary catheter after operation [ (6.8±1.5) vs (7.1±2.0) d ], the retention rate of postoperative erectile function (31.0% vs 27.0%), the satisfaction rate of immediate postoperative urine control (79.3% vs 75.6%), the positive rate of postoperative lymph node pathology (17.2% vs 14.6%), the length of incision [ (5.1±0.5) vs (6.1±0.4) cm ], the rate of PSA lower than 0.2 μg/L at 6 weeks after operation (86.2% vs 83.0%) of the two groups was not statistically significant(all P>0.05). The operation cost of group A[(62 000±4 000) yuan]was lower than group B[(68 000±4 000) yuan] (P<0.05). Conclusion: Extraperitoneal non-special channel device single-incision RARP is safe and feasible.

Type 2 diabetes is a global health burden driven by genetic and environmental factors. Continuous glucose monitoring (CGM) can effectively guide lifestyle interventions in non-diabetic. However, predefined CGM metrics fail to fully capture the dysglycemic information contained in the highdimensional time-series CGM data. This study employed deep learning to learn dysglycemia features from CGM data associated with diabetes and derived a digital biomarker of dysglycemia, validated against traditional dysglycemic biomarkers and diabetes polygenic risk score (PRS). Output of the deep learning model, called the deep learning-score, was significantly associated with multiple existing dysglycemic biomarkers and PRS of diabetes (P = 0.007). Moreover, existing CGM metrics were not associated with prevalent diabetes after adjusting for the deep learning-score, while the deep learning-score remained significantly associated with prevalent diabetes (P < 0.001) in a regression analysis. This digital biomarker demonstrated potential for providing dynamic feedback on dysglycemia and improving long-term intervention adherence.