Implementation of five machine learning methods to predict the 52-week blood glucose level in patients with type 2 diabetes.

Objective: For the patients who are suffering from type 2 diabetes, blood glucose level could be affected by multiple factors. An accurate estimation of the trajectory of blood glucose is crucial in clinical decision making. Frequent glucose measurement serves as a good source of data to train machine learning models for prediction purposes. This study aimed at using machine learning methods to predict blood glucose for type 2 diabetic patients. We investigated various parameters influencing blood glucose, as well as determined the most effective machine learning algorithm in predicting blood glucose. Patients and methods: 273 patients were recruited in this research. Several parameters such as age, diet, family history, BMI, alcohol intake, smoking status et al were analyzed. Patients who had glycosylated hemoglobin less than 6.5% after 52 weeks were considered as having achieved glycemic control and the rest as not achieving it. Five machine learning methods (KNN algorithm, logistic regression algorithm, random forest algorithm, support vector machine, and XGBoost algorithm) were compared to evaluate their performances in prediction accuracy. R 3.6.3 and Python 3.12 were used in data analysis. Results: The statistical variables for which p< 0.05 was obtained were BMI, pulse, Na, Cl, AKP. Compared with the other four algorithms, XGBoost algorithm has the highest accuracy (Accuracy=99.54% in training set and 78.18% in testing set) and AUC values (1.0 in training set and 0.68 in testing set), thus it is recommended to be used for prediction in clinical practice. Conclusion: When it comes to future blood glucose level prediction using machine learning methods, XGBoost algorithm scores the highest in effectiveness. This algorithm could be applied to assist clinical decision making, as well as guide the lifestyle of diabetic patients, in pursuit of minimizing risks of hyperglycemic or hypoglycemic events.

Cerebral Reorganization in Subacute Stroke Survivors after Virtual Reality-Based Training: A Preliminary Study.

BACKGROUND: Functional magnetic resonance imaging (fMRI) is a promising method for quantifying brain recovery and investigating the intervention-induced changes in corticomotor excitability after stroke. This study aimed to evaluate cortical reorganization subsequent to virtual reality-enhanced treadmill (VRET) training in subacute stroke survivors. METHODS: Eight participants with ischemic stroke underwent VRET for 5 sections per week and for 3 weeks. fMRI was conducted to quantify the activity of selected brain regions when the subject performed ankle dorsiflexion. Gait speed and clinical scales were also measured before and after intervention. RESULTS: Increased activation in the primary sensorimotor cortex of the lesioned hemisphere and supplementary motor areas of both sides for the paretic foot (p < 0.01) was observed postintervention. Statistically significant improvements were observed in gait velocity (p < 0.05). The change in voxel counts in the primary sensorimotor cortex of the lesioned hemisphere is significantly correlated with improvement of 10 m walk time after VRET (r = -0.719). CONCLUSIONS: We observed improved walking and increased activation in cortical regions of stroke survivors after VRET training. Moreover, the cortical recruitment was associated with better walking function. Our study suggests that cortical networks could be a site of plasticity, and their recruitment may be one mechanism of training-induced recovery of gait function in stroke. This trial is registered with ChiCTR-IOC-15006064.

Mechanism of Kinect-based virtual reality training for motor functional recovery of upper limbs after subacute stroke.

The Kinect-based virtual reality system for the Xbox 360 enables users to control and interact with the game console without the need to touch a game controller, and provides rehabilitation training for stroke patients with lower limb dysfunctions. However, the underlying mechanism remains unclear. In this study, 18 healthy subjects and five patients after subacute stroke were included. The five patients were scanned using functional MRI prior to training, 3 weeks after training and at a 12-week follow-up, and then compared with healthy subjects. The Fugl-Meyer Assessment and Wolf Motor Function Test scores of the hemiplegic upper limbs of stroke patients were significantly increased 3 weeks after training and at the 12-week follow-up. Functional MRI results showed that contralateral primary sensorimotor cortex was activated after Kinect-based virtual reality training in the stroke patients compared with the healthy subjects. Contralateral primary sensorimotor cortex, the bilateral supplementary motor area and the ipsilateral cerebellum were also activated during hand-clenching in all 18 healthy subjects. Our findings indicate that Kinect-based virtual reality training could promote the recovery of upper limb motor function in subacute stroke patients, and brain reorganization by Kinect-based virtual reality training may be linked to the contralateral sensorimotor cortex.