(1) Background: the use of machine learning techniques for the purpose of anticipating hypoglycemia has increased considerably in the past few years. Hypoglycemia is the drop in blood glucose below ...critical levels in diabetic patients. This may cause loss of cognitive ability, seizures, and in extreme cases, death. In almost half of all the severe cases, hypoglycemia arrives unannounced and is essentially asymptomatic. The inability of a diabetic patient to anticipate and intervene the occurrence of a hypoglycemic event often results in crisis. Hence, the prediction of hypoglycemia is a vital step in improving the life quality of a diabetic patient. The objective of this paper is to review work performed in the domain of hypoglycemia prediction by using machine learning and also to explore the latest trends and challenges that the researchers face in this area; (2) Methods: literature obtained from PubMed and Google Scholar was reviewed. Manuscripts from the last five years were searched for this purpose. A total of 903 papers were initially selected of which 57 papers were eventually shortlisted for detailed review; (3) Results: a thorough dissection of the shortlisted manuscripts provided an interesting split between the works based on two categories: hypoglycemia prediction and hypoglycemia detection. The entire review was carried out keeping this categorical distinction in perspective while providing a thorough overview of the machine learning approaches used to anticipate hypoglycemia, the type of training data, and the prediction horizon.
In this paper, we present a methodology based on generative adversarial network architecture to generate synthetic data sets with the intention of augmenting continuous glucose monitor data from ...individual patients. We use these synthetic data with the aim of improving the overall performance of prediction models based on machine learning techniques. Experiments were performed on two cohorts of patients suffering from type 1 diabetes mellitus with significant differences in their clinical outcomes. In the first contribution, we have demonstrated that the chosen methodology is able to replicate the intrinsic characteristics of individual patients following the statistical distributions of the original data. Next, a second contribution demonstrates the potential of synthetic data to improve the performance of machine learning approaches by testing and comparing different prediction models for the problem of predicting nocturnal hypoglycemic events in type 1 diabetic patients. The results obtained for both generative and predictive models are quite encouraging and set a precedent in the use of generative techniques to train new machine learning models.
A number of bifidobacterial species are found at a particularly high prevalence and abundance in faecal samples of healthy breastfed infants, a phenomenon that is believed to be, at least partially, ...due to the ability of bifidobacteria to metabolize Human Milk Oligosaccharides (HMOs). In the current study, we isolated a novel strain of Bifidobacterium kashiwanohense, named APCKJ1, from the faeces of a four-week old breastfed infant, based on the ability of the strain to utilise the HMO component fucosyllactose. We then determined the full genome sequence of this strain, and employed the generated data to analyze fucosyllactose metabolism in B. kashiwanohense APCKJ1. Transcriptomic and growth analyses, combined with metabolite analysis, in vitro hydrolysis assays and heterologous expression, allowed us to elucidate the pathway for fucosyllactose metabolism in B. kashiwanohense APCKJ1. Homologs of the key genes for this metabolic pathway were identified in particular in infant-derived members of the Bifdobacterium genus, revealing the apparent niche-specific nature of this pathway, and allowing a broad perspective on bifidobacterial fucosyllactose and L-fucose metabolism.
In this paper, we present solution algorithms for the cycle hub location problem (CHLP), which seeks to locate
p
hub facilities that are connected by means of a cycle, and to assign non-hub nodes to ...hubs so as to minimize the total cost of routing flows through the network. This problem is useful in modeling applications in transportation and telecommunications systems, where large setup costs on the links and reliability requirements make cycle topologies a prominent network architecture. We present a branch-and-cut algorithm that uses a flow-based formulation and two families of mixed-dicut inequalities as a lower bounding procedure at nodes of the enumeration tree. We also introduce a metaheuristic based on greedy randomized adaptive search procedure to obtain initial upper bounds for the exact algorithm and to obtain feasible solutions for large-scale instances of the CHLP. Numerical results on a set of benchmark instances with up to 100 nodes and 8 hubs confirm the efficiency of the proposed solution algorithms.
Artificial intelligence methods in combination with the latest technologies, including medical devices, mobile computing, and sensor technologies, have the potential to enable the creation and ...delivery of better management services to deal with chronic diseases. One of the most lethal and prevalent chronic diseases is diabetes mellitus, which is characterized by dysfunction of glucose homeostasis.
The objective of this paper is to review recent efforts to use artificial intelligence techniques to assist in the management of diabetes, along with the associated challenges.
A review of the literature was conducted using PubMed and related bibliographic resources. Analyses of the literature from 2010 to 2018 yielded 1849 pertinent articles, of which we selected 141 for detailed review.
We propose a functional taxonomy for diabetes management and artificial intelligence. Additionally, a detailed analysis of each subject category was performed using related key outcomes. This approach revealed that the experiments and studies reviewed yielded encouraging results.
We obtained evidence of an acceleration of research activity aimed at developing artificial intelligence-powered tools for prediction and prevention of complications associated with diabetes. Our results indicate that artificial intelligence methods are being progressively established as suitable for use in clinical daily practice, as well as for the self-management of diabetes. Consequently, these methods provide powerful tools for improving patients' quality of life.
(1) Background: nocturnal hypoglycemia (NH) is one of the most challenging side effects of multiple doses of insulin (MDI) therapy in type 1 diabetes (T1D). This work aimed to investigate the ...feasibility of a machine-learning-based prediction model to anticipate NH in T1D patients on MDI. (2) Methods: ten T1D adults were studied during 12 weeks. Information regarding T1D management, continuous glucose monitoring (CGM), and from a physical activity tracker were obtained under free-living conditions at home. Supervised machine-learning algorithms were applied to the data, and prediction models were created to forecast the occurrence of NH. Individualized prediction models were generated using multilayer perceptron (MLP) and a support vector machine (SVM). (3) Results: population outcomes indicated that more than 70% of the NH may be avoided with the proposed methodology. The predictions performed by the SVM achieved the best population outcomes, with a sensitivity and specificity of 78.75% and 82.15%, respectively. (4) Conclusions: our study supports the feasibility of using ML techniques to address the prediction of nocturnal hypoglycemia in the daily life of patients with T1D on MDI, using CGM and a physical activity tracker.
NMR spectroscopy is a mainstay of metabolic profiling approaches to investigation of physiological and pathological processes. The one‐dimensional proton pulse sequences typically used in phenotyping ...large numbers of samples generate spectra that are rich in information but where metabolite identification is often compromised by peak overlap. Recently developed pure shift (PS) NMR spectroscopy, where all J‐coupling multiplicities are removed from the spectra, has the potential to simplify the complex proton NMR spectra that arise from biosamples and hence to aid metabolite identification. Here we have evaluated two complementary approaches to spectral simplification: the HOBS (band‐selective with real‐time acquisition) and the PSYCHE (broadband with pseudo‐2D interferogram acquisition) pulse sequences. We compare their relative sensitivities and robustness for deconvolving both urine and serum matrices. Both methods improve resolution of resonances ranging from doublets, triplets and quartets to more complex signals such as doublets of doublets and multiplets in highly overcrowded spectral regions. HOBS is the more sensitive method and takes less time to acquire in comparison with PSYCHE, but can introduce unavoidable artefacts from metabolites with strong couplings, whereas PSYCHE is more adaptable to these types of spin system, although at the expense of sensitivity. Both methods are robust and easy to implement. We also demonstrate that strong coupling artefacts contain latent connectivity information that can be used to enhance metabolite identification. Metabolite identification is a bottleneck in metabolic profiling studies. In the case of NMR, PS experiments can be included in metabolite identification workflows, providing additional capability for biomarker discovery.
HOBS and PSYCHE are robust and easy to implement pure shift (PS) experiments to identify metabolites in urine and serum. Strong coupling artefacts contain latent connectivity information that can be used to enhance metabolite identification. PS experiments can be included in metabolite identification workflows, providing additional capability for biomarker discovery.
An understanding of the metabolic determinants of postexercise appetite regulation would facilitate development of adjunctive therapeutics to suppress compensatory eating behaviours and improve the ...efficacy of exercise as a weight‐loss treatment. Metabolic responses to acute exercise are, however, dependent on pre‐exercise nutritional practices, including carbohydrate intake. We therefore aimed to determine the interactive effects of dietary carbohydrate and exercise on plasma hormonal and metabolite responses and explore mediators of exercise‐induced changes in appetite regulation across nutritional states. In this randomized crossover study, participants completed four 120 min visits: (i) control (water) followed by rest; (ii) control followed by exercise (30 min at ∼75% of maximal oxygen uptake); (iii) carbohydrate (75 g maltodextrin) followed by rest; and (iv) carbohydrate followed by exercise. An ad libitum meal was provided at the end of each 120 min visit, with blood sample collection and appetite assessment performed at predefined intervals. We found that dietary carbohydrate and exercise exerted independent effects on the hormones glucagon‐like peptide 1 (carbohydrate, 16.8 pmol/L; exercise, 7.4 pmol/L), ghrelin (carbohydrate, −48.8 pmol/L; exercise: −22.7 pmol/L) and glucagon (carbohydrate, 9.8 ng/L; exercise, 8.2 ng/L) that were linked to the generation of distinct plasma 1H nuclear magnetic resonance metabolic phenotypes. These metabolic responses were associated with changes in appetite and energy intake, and plasma acetate and succinate were subsequently identified as potential novel mediators of exercise‐induced appetite and energy intake responses. In summary, dietary carbohydrate and exercise independently influence gastrointestinal hormones associated with appetite regulation. Future work is warranted to probe the mechanistic importance of plasma acetate and succinate in postexercise appetite regulation.
Key points
Carbohydrate and exercise independently influence key appetite‐regulating hormones.
Temporal changes in postexercise appetite are linked to acetate, lactate and peptide YY.
Postexercise energy intake is associated with glucagon‐like peptide 1 and succinate levels.
figure legend Our work aimed to explore hormonal and metabolic mediators of exercise‐induced changes in appetite and energy intake across nutritional states. Twelve male participants completed four study visits involving intake of water (control) or carbohydrate with a 30 min rest or high‐intensity exercise session. Plasma samples were collected throughout the 120 min study periods to quantify gastrointestinal hormone release and 1H nuclear magnetic resonance metabolite profiles. Visual analog scales were used to investigate appetite responses, and an ad libitum meal was provided at the end of each study visit to evaluate energy intake. Temporal changes in acetate, lactate and peptide YY were associated with supressed appetite responses in both exercise conditions. A consistent negative association between glucagon‐like peptide 1 and succinate levels with meal energy intake was found in both exercise conditions.
Summary
Background
The gastrointestinal microbiota has an important role in mucosal immune homoeostasis and may contribute to maintaining mucosal healing in Crohn's disease (CD).
Aim
To identify ...changes in the microbiota, metabolome and protease activity associated with mucosal healing in established paediatric CD
Methods
Twenty‐five participants aged 3‐18 years with CD, disease duration of over 6 months, and maintenance treatment with biological therapy were recruited. They were divided into a low calprotectin group (faecal calprotectin <100 μg/g, “mucosal healing,” n = 11), and a high calprotectin group (faecal calprotectin >100 μg/g, “mucosal inflammation,” n = 11). 16S gene‐based metataxonomics, 1H‐NMR spectroscopy‐based metabolic profiling and protease activity assays were performed on stool samples.
Results
Relative abundance of Dialister species was six‐times greater in the low calprotectin group (q = 0.00999). Alpha and beta diversity, total protease activity and inferred metagenomic profiles did not differ between groups. Pentanoate (valerate) and lysine were principal discriminators in a machine‐learning model which differentiated high and low calprotectin samples using NMR spectra (R2 0.87, Q2 0.41). Mean relative concentration of pentanoate was 1.35‐times greater in the low calprotectin group (95% CI 1.03‐1.68, P = 0.036) and was positively correlated with Dialister. Mean relative concentration of lysine was 1.54‐times greater in the high calprotectin group (95% CI 1.05‐2.03, P = 0.028).
Conclusions
This multiomic study identified an increase in Dialister species and pentanoate, and a decrease in lysine, in patients with “mucosal healing.” It supports further investigation of these as potential novel therapeutic targets in CD.
•Risk-based postprandial hypoglycemia prediction is feasible for patients with Type 1 Diabetes using machine-learning techniques.•A high sensitivity and low false positive rate was obtained for Level ...1 and Level 2 hypoglycemia using our methodology.•More than two thirds of the hypoglycemic events could be avoided thanks to our method.•The methodology can be easily integrated in platforms based on continuous glucose monitoring and intensive insulin management.
Predicting insulin-induced postprandial hypoglycemic events is critical for the safety of type 1 diabetes patients because an early warning of hypoglycemia facilitates correction of the insulin bolus before its administration. The postprandial hypoglycemic event counts can be lowered by reducing the size of the bolus based on a reliable prediction but at the cost of increasing the average blood glucose.
We developed a method for predicting postprandial hypoglycemia using machine learning techniques personalized to each patient. The proposed system enables on-line therapeutic decision making for patients using a sensor augmented pump therapy. Two risk-based approaches were developed for a window of 240 min after the meal/bolus, and they were tested based on real retrospective data from 10 patients using 70 mg/dL and 54 mg/dL as thresholds according to the consensus for Level 1 and Level 2 hypoglycemia, respectively. Due to the small size of the patient cohort, we trained personalized models for each patient.
The median specificity and sensitivity were 79% and 71% for Level 1 hypoglycemia, respectively, and 81% and 77% for Level 2.
The results demonstrated that it is feasible to anticipate hypoglycemic events with a reasonable false-positive rate. The accuracy of the results and the trade-off between performance metrics allow its use in decision support systems for patients who wear insulin pumps.
PDF
