Explainable Machine Learning in the Prediction of Depression DOI Creative Commons

Christina Mimikou,

Christos Kokkotis, Dimitrios Tsiptsios

et al.

Diagnostics, Journal Year: 2025, Volume and Issue: 15(11), P. 1412 - 1412

Published: June 2, 2025

Background: Depression constitutes a major public health issue, being one of the leading causes burden disease worldwide. The risk depression is determined by both genetic and environmental factors. While factors cannot be altered, identification potentially reversible crucial in order to try limit prevalence depression. Aim: A cross-sectional, questionnaire-based study on sample from multicultural region Thrace northeast Greece was designed assess potential association with several sociodemographic characteristics, lifestyle, status. employed four machine learning (ML) methods depression: logistic regression (LR), support vector (SVM), XGBoost, neural networks (NNs). These models were compared identify best-performing approach. Additionally, algorithm (GA) utilized for feature selection SHAP (SHapley Additive exPlanations) interpreting contributions each feature. Results: XGBoost classifier demonstrated highest performance test dataset predict excellent accuracy (97.83%), NNs close second (accuracy, 97.02%). 15 most significant identified GA algorithm. analysis revealed that anxiety, education level, alcohol consumption, body mass index influential predictors Conclusions: findings provide valuable insights development personalized interventions clinical strategies, ultimately promoting improved mental well-being individuals. Future research should expand datasets enhance model accuracy, enabling early detection healthcare systems better intervention.

Language: Английский

Artificial Intelligence in Primary Care Decision-Making: Survey of Healthcare Professionals in Saudi Arabia DOI Open Access

Najlaa Mohammad Alsudairy,

Alaa Omar Alahdal,

Mohammed Alrashidi

et al.

Cureus, Journal Year: 2025, Volume and Issue: unknown

Published: April 9, 2025

Artificial intelligence (AI) has the potential to revolutionize healthcare, particularly in primary care, by improving clinical decision-making and patient outcomes. AI technologies, such as machine learning natural language processing, can assist clinicians diagnosing conditions, predicting outcomes, recommending treatments, identifying at-risk individuals. Despite its potential, adoption care is slow due various challenges, including resource limitations, clinician training, concerns about reliability of systems. Understanding healthcare professionals' perceptions crucial for overcoming these barriers promoting integration into practice. A cross-sectional, survey-based study was conducted assess awareness, usage, perceptions, Saudi Arabia. The included 250 professionals from settings across urban, rural, hospital-based clinics. Data were collected via an electronic survey that both quantitative qualitative questions, analyzed using descriptive inferential statistics. total participated survey. majority physicians (44.8%), with remaining participants consisting nurses (27.2%), medical assistants (15.6%), administrators (8.8%). Awareness tools mixed, 14.8% respondents very familiar 47.2% unfamiliar. Thirty-one percent reported tools, primarily diagnostic support (59.5%). Common high implementation costs (49.2%) lack training (34%). significant portion (48%) expressed undermining human touch healthcare. hindered low familiarity well several barriers, cost, However, there optimism AI's decision-making. Overcoming through targeted education, infrastructure investment, further research essential realizing benefits

Language: Английский

Citations

0

Case Report: The intersection of psychiatry and medicine: diagnostic and ethical insights from case studies DOI Creative Commons
Francesco Monaco,

Annarita Vignapiano,

Martina D’Angelo

et al.

Frontiers in Psychiatry, Journal Year: 2025, Volume and Issue: 16

Published: April 22, 2025

The intersection of psychiatry and medicine presents unique diagnostic ethical challenges, particularly for conditions involving significant brain-body interactions, such as psychosomatic, somatopsychic, complex systemic disorders. This article explores the historical contemporary issues in diagnosing conditions, emphasizing fragmentation medical psychiatric knowledge, biases clinical guidelines, mismanagement illnesses. Diagnostic errors often arise from insufficient integration between general psychiatry, compounded by reliance on population-based guidelines that neglect individual patient needs. Misclassification like myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), Lyme disease, fibromyalgia psychosomatic or psychogenic has led to stigmatization delayed care. While these are referenced emblematic examples misclassified poorly understood disorders, five cases discussed this do not directly illustrate diseases. Instead, they exemplify shared dilemmas at medicine-psychiatry interface, including uncertainty, fragmentation, risk epistemic injustice. critically examines terms medically unexplained symptoms functional highlighting their limitations potential misuse. Case underscore consequences inaccuracies urgent need improved approaches. Ethical considerations also explored, respecting experiences, promoting individualized care, acknowledging inherent uncertainties diagnosis. Advances technologies brain imaging molecular diagnostics offer hope bridging gap medicine, enabling more accurate assessments better outcomes. concludes advocating comprehensive training interface a patient-centered approach integrates observation, research insights, nuanced understanding mind-body dynamics.

Language: Английский

Citations

0

Explainable Machine Learning in the Prediction of Depression DOI Creative Commons

Christina Mimikou,

Christos Kokkotis, Dimitrios Tsiptsios

et al.

Diagnostics, Journal Year: 2025, Volume and Issue: 15(11), P. 1412 - 1412

Published: June 2, 2025

Background: Depression constitutes a major public health issue, being one of the leading causes burden disease worldwide. The risk depression is determined by both genetic and environmental factors. While factors cannot be altered, identification potentially reversible crucial in order to try limit prevalence depression. Aim: A cross-sectional, questionnaire-based study on sample from multicultural region Thrace northeast Greece was designed assess potential association with several sociodemographic characteristics, lifestyle, status. employed four machine learning (ML) methods depression: logistic regression (LR), support vector (SVM), XGBoost, neural networks (NNs). These models were compared identify best-performing approach. Additionally, algorithm (GA) utilized for feature selection SHAP (SHapley Additive exPlanations) interpreting contributions each feature. Results: XGBoost classifier demonstrated highest performance test dataset predict excellent accuracy (97.83%), NNs close second (accuracy, 97.02%). 15 most significant identified GA algorithm. analysis revealed that anxiety, education level, alcohol consumption, body mass index influential predictors Conclusions: findings provide valuable insights development personalized interventions clinical strategies, ultimately promoting improved mental well-being individuals. Future research should expand datasets enhance model accuracy, enabling early detection healthcare systems better intervention.

Language: Английский

Citations

0