Objective: We aimed to explore the associations between common genetic risk variants with gestational diabetes mellitus (GDM) risk in Russian women and to assess their utility in the identification of GDM cases.Methods: We conducted a case-control study including 1,142 pregnant women (688 GDM cases and 454 controls) enrolled at Almazov National Medical Research Centre. The International Association of Diabetes and Pregnancy Study Groups criteria were used to diagnose GDM. A total of 11 single- nucleotide polymorphisms (SNPs), including those in HKDC1 (rs10762264), GCK (rs1799884), MTNR1B (rs10830963 and rs1387153), TCF7L2 (rs7903146 and rs12255372), KCNJ11 (rs5219), IGF2BP2 (rs4402960), IRS1 (rs1801278), FTO (rs9939609), and CDKAL1 (rs7754840) were genotyped using Taqman assays. A logistic regression model was used to calculate odds ratios (ORs) and their confidence intervals (CIs). A simple-count genetic risk score (GRS) was calculated using 6 SNPs. The area under the receiver operating characteristic curve (c-statistic) was calculated for the logistic regression model predicting the risk of GDM using clinical covariates, SNPs that had shown a significant association with GDM in our study, GRS, and their combinations.Results: Two variants in MTNR1B (rs1387153 and rs10830963) demonstrated a significant association with an increased risk of GDM. The association remained significant after adjustment for age, pre-gestational BMI, arterial hypertension, GDM in history, impaired glucose tolerance, polycystic ovary syndrome, family history of diabetes, and parity (P = 0.001 and P < 0.001, respectively). After being conditioned by each other, the effect of rs1387153 on GDM predisposition weakened while the effect of rs10830963 remained significant (P = 0.004). The risk of GDM was predicted by clinical variables (c-statistic 0.712, 95 % CI: 0.675 – 0.749), and the accuracy of prediction was modestly improved by adding GRS to the model (0.719, 95 % CI 0.682 – 0.755), and more by adding only rs10830963 (0.729, 95 % CI 0.693 – 0.764).Conclusion: Among 11 SNPs associated with T2D and/or GDM in other populations, we confirmed significant association with GDM for two variants in MTNR1B in Russian women. However, these variants showed limited value in the identification of GDM cases.

Postprandial blood glucose prediction is a crucial part of diabetes management. Recently, this topic has been of great interest, resulting in many research projects and published papers. Although different input parameters that might be beneficial for blood glucose prediction models were comprehensively discussed, specific data preprocessing, feature engineering and model tuning steps were not explained in detail in many of these papers. In this work, we developed and comprehensively described a data-driven blood glucose model based on a decision tree gradient boosting algorithm to predict different characteristics of postprandial glycemic responses; the model utilized meal-related data derived from a mobile app diary (including information on the glycemic index), food context (information on previous meals), characteristics of the individual patients and patient behavioral questionnaires. A set of rules was defined and implemented to detect incorrect meal records and to filter faulty data, and analyses were conducted on the overall food diary data and in particular, the data on the current meal for which the postprandial blood glucose response was calculated. Different gradient boosting models were trained and evaluated with parameters selected via random search cross-validation. The best models for the prediction of the incremental area under the blood glucose curve two hours after food intake had the following characteristics: R=0.631, MAE=0.373 mmol/L*h for the model not using data on current blood glucose; R=0.644, MAE=0.371 mmol/L*h for the model using data on the current blood glucose levels; and R=0.704, MAE=0.341 mmol/L*h for the model utilizing data on the continuous blood glucose trends before the meal. The impact of features was evaluated using Shapley values. The meal glycemic load, amount of carbohydrates in the meal, type of meal (e.g., breakfast), amount of starch and amount of food consumed 6 hours before the current meal were the most important contributors in the models.

Mobile and tablet apps, ubiquitous and pervasive computing, wearable computing and domotics for health.

The article proves the concept that an automatic diagnosis of depression can be achieved using audio recordings of the individuals' voices. DAIC-WOZ database was used as a data source. Audio and textual data were preprocessed and converted to a set of optimized parameters for two models. Appropriate Deep Learning models to detect depression in the transcripts of the audio recordings and voice quality features, were utilized. We created a text analysis model on a word-level using Natural Language Processing (NLP) techniques, and a voice quality analysis model on tense to breathy dimension. The text analysis model made its best performance with an F1-score equal to 0.8 (0.42) for non-depressed (depressed) individuals, while the voice quality model scored 0.76 (0.38). As a result, we had two models that would be implemented in a system for the diagnosis of depression.

The incorporation of glycemic index (GI) and glycemic load (GL) is a promising way to improve the accuracy of postprandial glycemic response (PPGR) prediction for personalized treatment of gestational diabetes (GDM). Our aim was to assess the prediction accuracy for PPGR prediction models with and without GI data in women with GDM and healthy pregnant women. The GI values were sourced from University of Sydney’s database and assigned to a food database used in the mobile app DiaCompanion. Weekly continuous glucose monitoring (CGM) data for 124 pregnant women (90 GDM and 34 control) were analyzed together with records of 1489 food intakes. Pearson correlation (R) was used to quantify the accuracy of predicted PPGRs from the model relative to those obtained from CGM. The final model for incremental area under glucose curve (iAUC120) prediction chosen by stepwise multiple linear regression had an R of 0.705 when GI/GL was included among input variables and an R of 0.700 when GI/GL was not included. In linear regression with coefficients acquired using regularization methods, which was tested on the data of new patients, R was 0.584 for both models (with and without inclusion of GI/GL). In conclusion, the incorporation of GI and GL only slightly improved the accuracy of PPGR prediction models when used in remote monitoring.

In recent decades, much work has been implemented in heart rate (HR) analysis using electrocardiographic (ECG) signals. We propose that algorithms developed to calculate HR based on detected R-peaks using ECG can be applied to seismocardiographic (SCG) signals, as they utilize common knowledge regarding heart rhythm and its underlying physiology. We implemented the experimental framework with methods developed for ECG signal processing and peak detection to be applied and evaluated on SCGs. Furthermore, we assessed and chose the best from all combinations of 15 peak detection and 6 preprocessing methods from the literature on the CEBS dataset available on Physionet. We then collected experimental data in the lab experiment to measure the applicability of the best-selected technique to the real-world data; the abovementioned method showed high precision for signals recorded during sitting rest (HR difference between SCG and ECG: 0.12 ± 0.35 bpm) and a moderate precision for signals recorded with interfering physical activity—reading out a book loud (HR difference between SCG and ECG: 6.45 ± 3.01 bpm) when compared to the results derived from the state-of-the-art photoplethysmographic (PPG) methods described in the literature. The study shows that computationally simple preprocessing and peak detection techniques initially developed for ECG could be utilized as the basis for HR detection on SCG, although they can be further improved.