Deep learning for prediction of depressive symptoms in a large textual dataset

AbstractDepression is a common illness worldwide with potentially severe implications. Early identification of depressive symptoms is a crucial first step towards assessment, intervention, and relapse prevention. With an increase in data sets with relevance for depression, and the advancement of machine learning, there is a potential to develop intelligent systems to detect symptoms of depression in written material. This work proposes an efficient approach using Long Short-Term Memory (LSTM)-based Recurrent Neural Network (RNN) to identify texts describing self-perceived symptoms of depression. The approach is applied on a large dataset from a public online information channel for young people in Norway. The dataset consists of youth’s own text-based questions on this information channel. Features are then provided from a one-hot process on robust features extracted from the reflection of possible symptoms of depression pre-defined by medical and psychological experts. The features are better than conventional approaches, which are mostly based on the word frequencies (i.e., some topmost frequent words are chosen as features from the whole text dataset and applied to model the underlying events in any text message) rather than symptoms. Then, a deep learning approach is applied (i.e., RNN) to train the time-sequential features discriminating texts describing depression symptoms from posts with no such descriptions (non-depression posts). Finally, the trained RNN is used to automatically predict depression posts. The system is compared against conventional approaches where it achieved superior performance than others. The linear discriminant space clearly reveals the robustness of the features by generating better clustering than other traditional features. Besides, since the features are based on the possible symptoms of depression, the system may generate meaningful explanations of the decision from machine learning models using an explainable Artificial Intelligence (XAI) algorithm called Local Interpretable Model-Agnostic Explanations (LIME). The proposed depression symptom feature-based approach shows superior performance compared to the traditional general word frequency-based approaches where frequency of the features gets more importance than the specific symptoms of depression. Although the proposed approach is applied on a Norwegian dataset, a similar robust approach can be applied on other depression datasets developed in other languages with proper annotations and symptom-based feature extraction. Thus, the depression prediction approach can be adopted to contribute to develop better mental health care technologies such as intelligent chatbots.

Download Full-text

Bereavement or Breakup: A Network Approach to Modeling Differential Pathways into Depression

10.31234/osf.io/38r4j ◽

2019 ◽

Author(s):

Julian Burger ◽

Margaret S. Stroebe ◽

Pasqualina Perrig-Chiello ◽

Henk A.W. Schut ◽

Stefanie Spahni ◽

...

Keyword(s):

Depressive Symptoms ◽

Graphical Models ◽

Contextual Factors ◽

Marital Disruption ◽

Depression Symptoms ◽

Depression Symptom ◽

Multiple Time ◽

Cross Sectional ◽

Network Analyses ◽

Within Subjects

Background: Prior network analyses demonstrated that the death of a loved one potentially precedes specific depression symptoms, primarily loneliness, which in turn links to other depressive symptoms. In this study, we extend prior research by comparing depression symptom network structures following two types of marital disruption: bereavement versus separation. Methods: We fitted two Gaussian Graphical Models to cross-sectional data from a Swiss survey of older persons (145 bereaved, 217 separated, and 362 married controls), and compared symptom levels across bereaved and separated individuals. Results: Separated compared to widowed individuals were more likely to perceive an unfriendly environment and oneself as a failure. Both types of marital disruption were linked primarily to loneliness, from where different relations emerged to other depressive symptoms. Amongst others, loneliness had a stronger connection to perceiving oneself as a failure in separated compared to widowed individuals. Conversely, loneliness had a stronger connection to getting going in widowed individuals. Limitations: Analyses are based on cross-sectional between-subjects data, and conclusions regarding dynamic processes on the within-subjects level remain putative. Further, some of the estimated parameters in the network exhibited overlapping confidence intervals and their order needs to be interpreted with care. Replications should thus aim for studies with multiple time points and larger samples. Conclusions: The findings of this study add to a growing body of literature indicating that depressive symptom patterns depend on contextual factors. If replicated on the within-subjects level, such findings have implications for setting up patient-tailored treatment approaches in dependence of contextual factors.

Download Full-text

Examining the Temporal Dynamics of Anxiety and Depressive Symptoms During a Therapist-Supported, Smartphone-Based Intervention for Major Depression (Preprint)

10.2196/preprints.25280 ◽

2020 ◽

Author(s):

Santiago Allende ◽

Valerie Forman-Hoffman ◽

Philippe Goldin

Keyword(s):

Depressive Symptoms ◽

Temporal Relationship ◽

Temporal Dynamics ◽

Digital Health ◽

Quadratic Function ◽

Depression Symptoms ◽

Symptoms Of Depression ◽

Effect Model ◽

End Of Treatment ◽

Highly Correlated

UNSTRUCTURED Background: Anxiety and depression symptoms are highly correlated in adults with depression; however, little is known about their interaction and temporal dynamics of change during treatment. Thus, the primary aim of this study was to examine the temporal dynamics of anxiety and depressive symptoms during a 12-week therapist-supported, smartphone-delivered digital health intervention for symptoms of depression and anxiety, the Meru Health Program (MHP). Method: A total of 290 participants from the MHP were included in the present analyses (age Mean = 39.64, SD = 10.25 years; 79% female; 54% self-reported psychotropic medication use). A variance components model was used to examine whether (1) reporting greater anxiety during the current week relative to anxiety reported in other weeks would be associated with greater reporting of depressive symptoms during the current week, while a time-varying effect model was used to examine whether, (2) consistent with findings reported by Wright et al. (2014), the temporal relationship between anxiety and depressive symptoms during the intervention would be expressed as a quadratic function marked by a weak association at baseline, followed by an increase to a peak before demonstrating a negligible decrease until the end of treatment. Results: In support of hypothesis 1, we found that reporting greater anxiety symptoms during the current week relative to other weeks was associated with greater depressive symptoms during the current week. Contrary to hypothesis 2, the temporal relationship between anxiety and depressive symptoms evidenced a recurring pattern, with the association increasing during the initial weeks, decreasing during mid-treatment and sharply increasing toward the end of treatment. Conclusions: The present findings demonstrate that anxiety and depressive symptoms overlap and fluctuate in concert during a smartphone-based intervention for anxiety and depressive symptoms. The present findings may warrant more refined intervention strategies specifically tailored to co-occurring patterns of change in symptoms.

Download Full-text

Deep Semi-Supervised Learning Improves Universal Peptide Identification of Shotgun Proteomics Data

10.1101/2020.11.12.380881 ◽

2020 ◽

Author(s):

John T. Halloran ◽

Gregor Urban ◽

David Rocke ◽

Pierre Baldi

Keyword(s):

Machine Learning ◽

Deep Learning ◽

Peptide Identification ◽

Shotgun Proteomics ◽

Database Search ◽

Supervised Machine Learning ◽

Superior Performance ◽

Support Vector ◽

Proteomics Data ◽

Learning Classifier

AbstractSemi-supervised machine learning post-processors critically improve peptide identification of shot-gun proteomics data. Such post-processors accept the peptide-spectrum matches (PSMs) and feature vectors resulting from a database search, train a machine learning classifier, and recalibrate PSMs using the trained parameters, often yielding significantly more identified peptides across q-value thresholds. However, current state-of-the-art post-processors rely on shallow machine learning methods, such as support vector machines. In contrast, the powerful training capabilities of deep learning models have displayed superior performance to shallow models in an ever-growing number of other fields. In this work, we show that deep models significantly improve the recalibration of PSMs compared to the most accurate and widely-used post-processors, such as Percolator and PeptideProphet. Furthermore, we show that deep learning is able to adaptively analyze complex datasets and features for more accurate universal post-processing, leading to both improved Prosit analysis and markedly better recalibration of recently developed database-search functions.

Download Full-text

FKBP5 Genotype Linked to Combined PTSD-Depression Symptom in Chinese Earthquake Survivors

The Canadian Journal of Psychiatry ◽

10.1177/0706743719870505 ◽

2019 ◽

Vol 64 (12) ◽

pp. 863-871

Author(s):

Gen Li ◽

Li Wang ◽

Kunlin Zhang ◽

Chengqi Cao ◽

Xing Cao ◽

...

Keyword(s):

Depressive Symptoms ◽

Genetic Basis ◽

Latent Profile Analysis ◽

Common Mental Disorders ◽

Phenotypic Heterogeneity ◽

Depression Symptoms ◽

Depression Symptom ◽

Environment Interaction ◽

Post Traumatic Stress ◽

2008 Wenchuan Earthquake

Background: Post-traumatic stress disorder (PTSD) and depression are common mental disorders in individuals experiencing traumatic events. To date, few studies have studied the relationship between genetic basis and phenotypic heterogeneity of traumatized individuals. The present study examined the effects of four FKBP5 SNPs (rs1360780, rs3800373, rs9296158, and rs9470080) in four postdisaster groups (low symptom, predominantly depressive, predominantly PTSD, and combined PTSD-depression symptom groups) as identified by latent profile analysis. Methods: A total of 1,140 adults who experienced the 2008 Wenchuan earthquake participated in our study. Earthquake-related trauma, PTSD, and depressive symptoms were measured using standard psychometric instruments. The four FKBP5 SNPs were genotyped using a custom-by-design 2 × 48-Plex SNP scan™ Kit. Results: After adjusting for covariates, the main and gene–environment interaction effects of rs9470080 were all significant when the combined PTSD-depression group was compared with the low symptoms, predominantly depression and predominantly PTSD groups. rs9470080 TT genotype carriers had a higher risk of developing high co-occurring PTSD and depression symptoms than the C allele carriers. However, when trauma exposure was severe, the TT genotype carriers and C allele carriers did not differ in the risk of developing high co-occurring PTSD and depressive symptoms. The other three SNPs demonstrated no significant effects. Moreover, the rs3800373-rs9296158-rs1360780-rs9470080 haplotype A-G-C-T was found significantly associated with combined PTSD-depression symptoms. Conclusion: Our findings support the genetic basis of phenotypic heterogeneity in people exposed to trauma. Furthermore, the results reveal the possibility that the variants of FKBP5 gene may be associated with depression-PTSD comorbidity.

Download Full-text

Eye tracking and attentional bias for depressive internet memes in depression

Experimental Brain Research ◽

10.1007/s00221-020-06001-8 ◽

2020 ◽

Author(s):

Umair Akram ◽

Jason G. Ellis ◽

Glhenda Cau ◽

Frayer Hershaw ◽

Ashlieen Rajenthran ◽

...

Keyword(s):

Depressive Symptoms ◽

Eye Movements ◽

Attentional Bias ◽

Control Group ◽

Depression Symptoms ◽

Significant Group ◽

Symptoms Of Depression ◽

Total Fixation ◽

Potential Benefits ◽

Type Interaction

AbstractPrevious research highlights the potential benefits of engaging with depressive internet memes for those experiencing symptoms of depression. This study aimed to determine whether: compared to non-depressed controls, individuals experiencing depressive symptoms were quicker to orient and maintain overall attention for internet memes depicting depressive content relative to neutral memes. N = 21 individuals were grouped based on the severity of reported depression symptoms using the PhQ-9. Specifically, a score of: ≤ 4 denoted the control group; and ≥ 15 the depressive symptoms group. Participants viewed a series of meme pairs depicting depressive and neutral memes for periods of 4000 ms. Data for the first fixation onset and duration, total fixation count and total fixation and gaze duration of eye-movements were recorded. A significant group x meme-type interaction indicated that participants with depressive symptoms displayed significantly more fixations on depressive rather than neutral memes. These outcomes provide suggestive evidence for the notion that depressive symptoms are associated with an attentional bias towards socio-emotionally salient stimuli.

Download Full-text

THE RELATIONSHIP BETWEEN MARITAL STATUS, COHORT, AND DEPRESSION SYMPTOMS IN TAIWAN

Innovation in Aging ◽

10.1093/geroni/igz038.1140 ◽

2019 ◽

Vol 3 (Supplement_1) ◽

pp. S311-S312

Author(s):

Fang-Yi Huang ◽

Min Li

Keyword(s):

Depressive Symptoms ◽

Marital Status ◽

Baby Boom ◽

Depression Symptoms ◽

Depression Symptom ◽

Economic Decline ◽

Logistic Regression Models ◽

Financial Decline ◽

Married Female ◽

The Relationship

Abstract Objectives: The relationship between marital status and depression symptoms is well documented. However, how the negative economic shock affect relationship differ by gender and cohort is still indecisive. The dataset “2011 wave of the Taiwan Longitudinal Study in Aging” and logistic regression models were used in the study. The results: Marital status is related to depression symptoms, but it differs by gendered cohort. With considering financial shock, there is no difference of depressive symptom between divorced and married female. The divorced and widowed have 4.81 and 2.47 times higher of getting depression symptom than the married for baby boom female. Being divorced is 3.67 times higher of getting depressive symptoms than being married for baby boom male. For WWII female, the widows are 1.78 times higher to have depressive symptoms than the married. being divorced, widowers, and single are 3.32, 2.21 and 2.90 times higher of getting depressive symptoms than being married for WWII male. Being divorced is 3.67 times higher of getting depressive symptoms than being married for baby boom male. In conclusions, people with unstable marital statuses are more depressed than the married. In particular, the effect of unstable marital statuses on depression could be account for by financial decline for women but not men. Given the policy emphasis on those with unstable marital status and economic decline, divorce female and single baby boom female may represent particular groups in whom interventions designed to financially support.

Download Full-text

Short-Term Forecasting of Photovoltaic Solar Power Production Using Variational Auto-Encoder Driven Deep Learning Approach

Applied Sciences ◽

10.3390/app10238400 ◽

2020 ◽

Vol 10 (23) ◽

pp. 8400 ◽

Cited By ~ 1

Author(s):

Abdelkader Dairi ◽

Fouzi Harrou ◽

Ying Sun ◽

Sofiane Khadraoui

Keyword(s):

Machine Learning ◽

Deep Learning ◽

Solar Power ◽

Power Production ◽

Superior Performance ◽

Support Vector ◽

Learning Models ◽

Short Term ◽

Learning Methods ◽

Short Term Forecasting

The accurate modeling and forecasting of the power output of photovoltaic (PV) systems are critical to efficiently managing their integration in smart grids, delivery, and storage. This paper intends to provide efficient short-term forecasting of solar power production using Variational AutoEncoder (VAE) model. Adopting the VAE-driven deep learning model is expected to improve forecasting accuracy because of its suitable performance in time-series modeling and flexible nonlinear approximation. Both single- and multi-step-ahead forecasts are investigated in this work. Data from two grid-connected plants (a 243 kW parking lot canopy array in the US and a 9 MW PV system in Algeria) are employed to show the investigated deep learning models’ performance. Specifically, the forecasting outputs of the proposed VAE-based forecasting method have been compared with seven deep learning methods, namely recurrent neural network, Long short-term memory (LSTM), Bidirectional LSTM, Convolutional LSTM network, Gated recurrent units, stacked autoencoder, and restricted Boltzmann machine, and two commonly used machine learning methods, namely logistic regression and support vector regression. The results of this investigation demonstrate the satisfying performance of deep learning techniques to forecast solar power and point out that the VAE consistently performed better than the other methods. Also, results confirmed the superior performance of deep learning models compared to the two considered baseline machine learning models.

Download Full-text

Frequency and heritability of depression symptomatology in the second half of life: evidence from Danish twins over 45

Psychological Medicine ◽

10.1017/s0033291702006207 ◽

2002 ◽

Vol 32 (7) ◽

pp. 1175-1185 ◽

Cited By ~ 36

Author(s):

W. JOHNSON ◽

M. McGUE ◽

D. GAIST ◽

J. W. VAUPEL ◽

K. CHRISTENSEN

Keyword(s):

Depressive Symptoms ◽

Age Groups ◽

Well Being ◽

The Elderly ◽

Depression Symptoms ◽

Same Sex ◽

Symptoms Of Depression ◽

Genetic Origins ◽

Depression Symptomatology ◽

Special Risk

Background. Self-reported depressive symptoms among the elderly have generated considerable interest because they are readily available measures of overall well-being in a population often thought to be at special risk for mental disorder.Method. The heritability of depression symptoms was investigated in a sample of 2169 pairs of Danish twins (1033 MZ and 1136 same sex DZ) ranging in age from 45 to over 95. Twins completed an interview assessment that identified symptoms of depression, which were scored on Affective, Somatic and Total scales.Results. Overall heritability estimates (a2) for the Affective (a2 = 0.27, (95% CI 0.22–0.32)). Somatic (a2 = 0.26, (0.21–0.32)), and Total (a2 = 0.29, (0.22–0.34)) scales were all moderate, statistically significant and similar to results from other studies. To assess possible variations in heritability across the wide age span, the sample was stratified into age groups in increments of 10 years. The magnitude of heritable influence did not vary significantly with age or sex. Somatic scale heritability tended to be greater for females than for males, though this difference was not statistically significant. The genetic correlation between the Affective and Somatic scales was 0.71, suggesting substantial common genetic origins.Conclusions. Though the frequency of self-reported depressive symptoms increased with age in this sample, their heritability did not.

Download Full-text

Applied Machine Learning for Spine Surgeons: Predicting Outcome for Patients Undergoing Treatment for Lumbar Disc Herniation Using PRO Data

Global Spine Journal ◽

10.1177/2192568220967643 ◽

2020 ◽

pp. 219256822096764

Author(s):

Casper Friis Pedersen ◽

Mikkel Østerheden Andersen ◽

Leah Yacat Carreon ◽

Søren Eiskjær

Keyword(s):

Machine Learning ◽

Deep Learning ◽

Lumbar Disc Herniation ◽

Disc Herniation ◽

Lumbar Disc ◽

Multivariate Adaptive Regression Splines ◽

Superior Performance ◽

Support Vector ◽

Conventional Models

Study Design: Retrospective/prospective study. Objective: Models based on preoperative factors can predict patients’ outcome at 1-year follow-up. This study measures the performance of several machine learning (ML) models and compares the results with conventional methods. Methods: Inclusion criteria were patients who had lumbar disc herniation (LDH) surgery, identified in the Danish national registry for spine surgery. Initial training of models included 16 independent variables, including demographics and presurgical patient-reported measures. Patients were grouped by reaching minimal clinically important difference or not for EuroQol, Oswestry Disability Index, Visual Analog Scale (VAS) Leg, and VAS Back and by their ability to return to work at 1 year follow-up. Data were randomly split into training, validation, and test sets by 50%/35%/15%. Deep learning, decision trees, random forest, boosted trees, and support vector machines model were trained, and for comparison, multivariate adaptive regression splines (MARS) and logistic regression models were used. Model fit was evaluated by inspecting area under the curve curves and performance during validation. Results: Seven models were arrived at. Classification errors were within ±1% to 4% SD across validation folds. ML did not yield superior performance compared with conventional models. MARS and deep learning performed consistently well. Discrepancy was greatest among VAS Leg models. Conclusions: Five predictive ML and 2 conventional models were developed, predicting improvement for LDH patients at the 1-year follow-up. We demonstrate that it is possible to build an ensemble of models with little effort as a starting point for further model optimization and selection.

Download Full-text

Wheat Lodging Detection from UAS Imagery Using Machine Learning Algorithms

Remote Sensing ◽

10.3390/rs12111838 ◽

2020 ◽

Vol 12 (11) ◽

pp. 1838 ◽

Cited By ~ 8

Author(s):

Zhao Zhang ◽

Paulo Flores ◽

C. Igathinathane ◽

Dayakar L. Naik ◽

Ravi Kiran ◽

...

Keyword(s):

Neural Network ◽

Machine Learning ◽

Deep Learning ◽

Standard Deviation ◽

Learning Algorithm ◽

Learning Algorithms ◽

Machine Learning Algorithms ◽

Superior Performance ◽

Support Vector ◽

Unmanned Aerial Systems

The current mainstream approach of using manual measurements and visual inspections for crop lodging detection is inefficient, time-consuming, and subjective. An innovative method for wheat lodging detection that can overcome or alleviate these shortcomings would be welcomed. This study proposed a systematic approach for wheat lodging detection in research plots (372 experimental plots), which consisted of using unmanned aerial systems (UAS) for aerial imagery acquisition, manual field evaluation, and machine learning algorithms to detect the occurrence or not of lodging. UAS imagery was collected on three different dates (23 and 30 July 2019, and 8 August 2019) after lodging occurred. Traditional machine learning and deep learning were evaluated and compared in this study in terms of classification accuracy and standard deviation. For traditional machine learning, five types of features (i.e. gray level co-occurrence matrix, local binary pattern, Gabor, intensity, and Hu-moment) were extracted and fed into three traditional machine learning algorithms (i.e., random forest (RF), neural network, and support vector machine) for detecting lodged plots. For the datasets on each imagery collection date, the accuracies of the three algorithms were not significantly different from each other. For any of the three algorithms, accuracies on the first and last date datasets had the lowest and highest values, respectively. Incorporating standard deviation as a measurement of performance robustness, RF was determined as the most satisfactory. Regarding deep learning, three different convolutional neural networks (simple convolutional neural network, VGG-16, and GoogLeNet) were tested. For any of the single date datasets, GoogLeNet consistently had superior performance over the other two methods. Further comparisons between RF and GoogLeNet demonstrated that the detection accuracies of the two methods were not significantly different from each other (p > 0.05); hence, the choice of any of the two would not affect the final detection accuracies. However, considering the fact that the average accuracy of GoogLeNet (93%) was larger than RF (91%), it was recommended to use GoogLeNet for wheat lodging detection. This research demonstrated that UAS RGB imagery, coupled with the GoogLeNet machine learning algorithm, can be a novel, reliable, objective, simple, low-cost, and effective (accuracy > 90%) tool for wheat lodging detection.

Download Full-text