Nevin Manimala

Artif Intell Med. 2021 Aug;118:102131. doi: 10.1016/j.artmed.2021.102131. Epub 2021 Jun 25.

ABSTRACT

Big data importance and potential are becoming more and more relevant nowadays, enhanced by the explosive growth of information volume that is being generated on the Internet in the last years. In this sense, many experts agree that social media networks are one of the internet areas with higher growth in recent years and one of the fields that are expected to have a more significant increment in the coming years. Similarly, social media sites are quickly becoming one of the most popular platforms to discuss health issues and exchange social support with others. In this context, this work presents a new methodology to process, classify, visualise and analyse the big data knowledge produced by the sociome on social media platforms. This work proposes a methodology that combines natural language processing techniques, ontology-based named entity recognition methods, machine learning algorithms and graph mining techniques to: (i) reduce the irrelevant messages by identifying and focusing the analysis only on individuals and patient experiences from the public discussion; (ii) reduce the lexical noise produced by the different ways in how users express themselves through the use of domain ontologies; (iii) infer the demographic data of the individuals through the combined analysis of textual, geographical and visual profile information; (iv) perform a community detection and evaluate the health topic study combining the semantic processing of the public discourse with knowledge graph representation techniques; and (v) gain information about the shared resources combining the social media statistics with the semantical analysis of the web contents. The practical relevance of the proposed methodology has been proven in the study of 1.1 million unique messages from >400,000 distinct users related to one of the most popular dietary fads that evolve into a multibillion-dollar industry, i.e., gluten-free food. Besides, this work analysed one of the least research fields studied on Twitter concerning public health (i.e., the allergies or immunology diseases as celiac disease), discovering a wide range of health-related conclusions.

PMID:34412847 | DOI:10.1016/j.artmed.2021.102131

Artif Intell Med. 2021 Aug;118:102120. doi: 10.1016/j.artmed.2021.102120. Epub 2021 May 28.

ABSTRACT

BACKGROUND AND AIM: Hypoglycaemia prediction play an important role in diabetes management being able to reduce the number of dangerous situations. Thus, it is relevant to present a systematic review on the currently available prediction algorithms and models for hypoglycaemia (or hypoglycemia in US English) prediction.

METHODS: This study aims to systematically review the literature on data-based algorithms and models using diabetics real data for hypoglycaemia prediction. Five electronic databases were screened for studies published from January 2014 to June 2020: ScienceDirect, IEEE Xplore, ACM Digital Library, SCOPUS, and PubMed.

RESULTS: Sixty-three eligible studies were retrieved that met the inclusion criteria. The review identifies the current trend in this topic: most of the studies perform short-term predictions (82.5%). Also, the review pinpoints the inputs and shows that information fusion is relevant for hypoglycaemia prediction. Regarding data-based models (80.9%) and hybrid models (19.1%) different predictive techniques are used: Artificial neural network (22.2%), ensemble learning (27.0%), supervised learning (20.6%), statistic/probabilistic (7.9%), autoregressive (7.9%), evolutionary (6.4%), deep learning (4.8%) and adaptative filter (3.2%). Artificial Neural networks and hybrid models show better results.

CONCLUSIONS: The data-based models for blood glucose and hypoglycaemia prediction should be able to provide a good balance between the applicability and performance, integrating complementary data from different sources or from different models. This review identifies trends and possible opportunities for research in this topic.

PMID:34412843 | DOI:10.1016/j.artmed.2021.102120

Artif Intell Med. 2021 Aug;118:102118. doi: 10.1016/j.artmed.2021.102118. Epub 2021 May 29.

ABSTRACT

Critical care clinicians are trained to analyze simultaneously multiple physiological parameters to predict critical conditions such as hemodynamic instability. We developed the Multi-task Learning Physiological Deep Learner (MTL-PDL), a deep learning algorithm that predicts simultaneously the mean arterial pressure (MAP) and the heart rate (HR). In an external validation dataset, our model exhibited very good calibration: R² of 0.747 (95% confidence interval, 0.692 to 0.794) and 0.850 (0.815 to 0.879) for respectively, MAP and HR prediction 60-minutes ahead of time. For acute hypotensive episodes defined as a MAP below 65 mmHg for 5 min, our MTL-PDL reached a predictive value of 90% for patients at very high risk (predicted MAP ≤ 60 mmHg) and 2‰ for patients at low risk (predicted MAP >70 mmHg). Based on its excellent prediction performance, the Physiological Deep Learner has the potential to help the clinician proactively adjust the treatment in order to avoid hypotensive episodes and end-organ hypoperfusion.

PMID:34412841 | DOI:10.1016/j.artmed.2021.102118

Ecol Lett. 2021 Aug 19. doi: 10.1111/ele.13862. Online ahead of print.

ABSTRACT

Predicting complex species-environment interactions is crucial for guiding conservation and mitigation strategies in a dynamically changing world. Phenotypic plasticity is a mechanism of trait variation that determines how individuals and populations adapt to changing and novel environments. For individuals, the effects of phenotypic plasticity can be quantified by measuring environment-trait relationships, but it is often difficult to predict how phenotypic plasticity affects populations. The assumption that environment-trait relationships validated for individuals indicate how populations respond to environmental change is commonly made without sufficient justification. Here we derive a novel general mathematical framework linking trait variation due to phenotypic plasticity to population dynamics. Applying the framework to the classical example of Nicholson’s blowflies, we show how seemingly sensible predictions made from environment-trait relationships do not generalise to population responses. As a consequence, trait-based analyses that do not incorporate population feedbacks risk mischaracterising the effect of environmental change on populations.

PMID:34412157 | DOI:10.1111/ele.13862

J Obstet Gynaecol Res. 2021 Aug 19. doi: 10.1111/jog.14993. Online ahead of print.

ABSTRACT

AIM: To determine whether preoperative genital hiatus at rest is predictive of medium-term prolapse recurrence.

METHODS: We conducted a retrospective study of women who underwent native tissue prolapse surgery from 2002 to 2017 with pelvic organ prolapse quantification data including resting genital hiatus at one of three time points: preoperatively, 6 weeks, and ≥1 year postoperatively. Demographics and clinical data were abstracted from the chart. Prolapse recurrence was defined by anatomic outcomes (Ba > 0, Bp > 0, and/or C ≥ -4) or retreatment. Descriptive statistics, bivariate analyses, and logistic regression analyses were performed.

RESULTS: Of the 165 women included, 36 (21.8%) had prolapse recurrence at an average of 1.5 years after surgery. Preoperative resting genital hiatus did not differ between women with surgical success versus recurrence (3.5 cm [interquartile range, IQR 2.25, 4.0) vs 3.5 cm (IQR 3.0, 4.0), p = 0.71). Point Bp was greater in the recurrence group at every time point. Preoperative Bp (odds ratio [OR] 1.24, confidence interval [CI] [1.06-1.45], p = 0.01) and days from surgery (OR 1.001, CI [1.000-1.001], p < 0.01) were independently associated with recurrence. Preoperative genital hiatus at rest and strain were significantly larger among women who underwent a colpoperineorrhaphy (rest: 4.0 [3.0, 4.5] cm vs 3.5 [3.0, 4.0] cm, p < 0.01; strain: 6.0 [4.0, 6.5] cm vs 5.0 [4.0, 6.0] cm, p = 0.01).

CONCLUSIONS: Preoperative genital hiatus at rest was not associated with prolapse recurrence when the majority of women underwent colpoperineorrhaphy. Preoperative Bp was more predictive of short-term prolapse recurrence. For every 1 cm increase in point Bp, there is a 24% increased odds of recurrence.

PMID:34412156 | DOI:10.1111/jog.14993

Stat Med. 2021 Aug 19. doi: 10.1002/sim.9167. Online ahead of print.

ABSTRACT

Statistical analysis of questionnaire data is often performed employing techniques from item-response theory. In this framework, it is possible to differentiate respondent profiles and characterize the questions (items) included in the questionnaire via interpretable parameters. These models are often crosssectional and aim at evaluating the performance of the respondents. The motivating application of this work is the analysis of psychometric questionnaires taken by a group of mothers at different time points and by their children at one later time point. The data are available through the GUSTO cohort study. To this end, we propose a Bayesian semiparametric model and extend the current literature by: (i) introducing temporal dependence among questionnaires taken at different time points; (ii) jointly modeling the responses to questionnaires taken from different, but related, groups of subjects (in our case mothers and children), introducing a further dependency structure and therefore sharing of information; (iii) allowing clustering of subjects based on their latent response profile. The proposed model is able to identify three main groups of mother/child pairs characterized by their response profiles. Furthermore, we report an interesting maternal reporting bias effect strongly affecting the clustering structure of the mother/child dyads.

PMID:34412151 | DOI:10.1002/sim.9167

Lasers Surg Med. 2021 Aug 19. doi: 10.1002/lsm.23453. Online ahead of print.

ABSTRACT

OBJECTIVES: Skin rejuvenation with radiofrequency has been a widely used treatment modality for the safe and efficient remodeling of the dermis and revision of textural irregularities, achieved with minimal downtime. The efficacy of fractional radiofrequency (FRF) specifically for acne scarring has not been widely established. The objective of this clinical trial was to establish the efficacy and safety of FRF for moderate to severe acne scarring in a wide range of Fitzpatrick skin types using two different applicator tips to deliver energy to the skin (80-pin of up to 124 mJ/pin and 160-pin of up to 62 mJ/pin).

METHODS: Enrolled subjects received a series of three FRF treatments to the full face, each 4 weeks apart. A visual analog scale was utilized to assess pain of the treatment. Subject satisfaction questionnaires were completed at follow-up visits at 6 and 12 weeks post final treatment. Photographs were graded for change by three blinded evaluators using the Global Aesthetic Improvement Scale (GAIS).

RESULTS: Image sets of 23 enrolled subjects were assessed by blinded evaluation, showing a statistically significant improvement (p = 0.009) from the baseline visit to the 12-week follow-up on the GAIS for acne scarring. Subject satisfaction was high with subjects giving an average satisfaction score of 3.27 (“satisfied”) out of 4. Pain was “mild” as treatments were rated an average of 2.15 on a 10-point visual analog scale. The GAIS score of the 80-pin tip improved patients’ acne scars treated with that applicator by 1.06 points and 0.85 for the 160-pin tip. Ninety-five percent (95.5%) of subjects reported either a mild, moderate, or significant improvement to their treatment area. Ninety-one percent of subjects reported that they would recommend the treatment to a friend.

CONCLUSION: FRF produced a statistically significant improvement in acne scarring when assessed by independent blinded evaluators. No serious adverse events resulted from treatment by either applicator tip. Treatment pain was low and tolerable among subjects of all Fitzpatrick skin types. Subjects had high levels of satisfaction with the results.

PMID:34412150 | DOI:10.1002/lsm.23453

Neural Comput. 2021 Aug 19;33(9):2473-2510. doi: 10.1162/neco_a_01416.

ABSTRACT

We investigate the use of parameterized families of information-theoretic measures to generalize the loss functions of generative adversarial networks (GANs) with the objective of improving performance. A new generator loss function, least kth-order GAN (LkGAN), is introduced, generalizing the least squares GANs (LSGANs) by using a kth-order absolute error distortion measure with k≥1 (which recovers the LSGAN loss function when k=2). It is shown that minimizing this generalized loss function under an (unconstrained) optimal discriminator is equivalent to minimizing the kth-order Pearson-Vajda divergence. Another novel GAN generator loss function is next proposed in terms of Rényi cross-entropy functionals with order α>0, α≠1. It is demonstrated that this Rényi-centric generalized loss function, which provably reduces to the original GAN loss function as α→1, preserves the equilibrium point satisfied by the original GAN based on the Jensen-Rényi divergence, a natural extension of the Jensen-Shannon divergence. Experimental results indicate that the proposed loss functions, applied to the MNIST and CelebA data sets, under both DCGAN and StyleGAN architectures, confer performance benefits by virtue of the extra degrees of freedom provided by the parameters k and α, respectively. More specifically, experiments show improvements with regard to the quality of the generated images as measured by the Fréchet inception distance score and training stability. While it was applied to GANs in this study, the proposed approach is generic and can be used in other applications of information theory to deep learning, for example, the issues of fairness or privacy in artificial intelligence.

PMID:34412112 | DOI:10.1162/neco_a_01416

PLoS One. 2021 Aug 19;16(8):e0255982. doi: 10.1371/journal.pone.0255982. eCollection 2021.

ABSTRACT

Milgram empirically showed that people knowing only connections to their friends could locate any person in the U.S. in a few steps. Later research showed that social network topology enables a node aware of its full routing to find an arbitrary target in even fewer steps. Yet, the success of people in forwarding efficiently knowing only personal connections is still not fully explained. To study this problem, we emulate it on a real location-based social network, Gowalla. It provides explicit information about friends and temporal locations of each user useful for studies of human mobility. Here, we use it to conduct a massive computational experiment to establish new necessary and sufficient conditions for achieving social search efficiency. The results demonstrate that only the distribution of friendship edges and the partial knowledge of friends of friends are essential and sufficient for the efficiency of social search. Surprisingly, the efficiency of the search using the original distribution of friendship edges is not dependent on how the nodes are distributed into space. Moreover, the effect of using a limited knowledge that each node possesses about friends of its friends is strongly nonlinear. We show that gains of such use grow statistically significantly only when this knowledge is limited to a small fraction of friends of friends.

PMID:34412110 | DOI:10.1371/journal.pone.0255982

J Clin Psychopharmacol. 2021 Aug 20. doi: 10.1097/JCP.0000000000001446. Online ahead of print.

ABSTRACT

PURPOSE/BACKGROUND: Antidepressants are among the most frequently prescribed medications during pregnancy and may affect fetal weight. Associations between antenatal antidepressant use and ultrasonographic measures of fetal development have rarely been examined. We hypothesized that the prescription of an antenatal antidepressant would be associated with lower estimated fetal weight (EFW).

METHODS/PROCEDURES: A retrospective analysis of routine ultrasonographic data extracted from electronic medical records was performed on a cohort of pregnant women with psychiatric diagnoses and grouped according to the presence of an antenatal antidepressant prescription (n = 32 antidepressant-prescribed and n = 44 antidepressant prescription-free). After stratifying for gestational age, comparisons included 13 ultrasonographic parameters, frequency of oligohydramnios and polyhydramnios and growth deceleration, and maternal serum protein markers assessed per routine care, including α-fetoprotein, free β-human chorionic gonadotropin, and unconjugated estriol levels, using t tests, nonparametric and Fisher tests, and effect sizes (ESs) were computed.

FINDINGS/RESULTS: No statistically significant EFW differences between groups at any time point were detected (P > 0.05). Antenatal antidepressant prescription was associated with lower femur length at weeks 33 to 40 (P = 0.046, ES = 0.75) and greater left ventricular diameter at weeks 25 to 32 (P = 0.04, ES = 1.18). No differences for frequency of oligohydramnios or polyhydramnios or growth deceleration were observed (P > 0.05). We did not detect group differences for maternal proteins (P > 0.05).

IMPLICATIONS/CONCLUSIONS: Our evidence suggested a lack of association between antenatal antidepressant prescription and lower EFW but indicated an association with lower femur length and greater left ventricular diameter in mid-late gestation. Future research should examine the clinical implications of these findings.

PMID:34412105 | DOI:10.1097/JCP.0000000000001446