JMIRx med最新文献

Background: Major depressive disorder (MDD) is a highly prevalent mental health condition with significant public health implications. Early detection is crucial for timely intervention, but current diagnostic methods often rely on subjective clinical assessments, leading to delayed or inaccurate diagnoses. Advances in neuroimaging and machine learning (ML) offer the potential for objective and accurate early detection.

Objective: This study aimed to develop and validate ML models using multisite functional magnetic resonance imaging data for the early detection of MDD, compare their performance, and evaluate their clinical applicability.

Methods: We used functional magnetic resonance imaging data from 1200 participants (600 with early-stage MDD and 600 healthy controls) across 3 public datasets. In total, 4 ML models-support vector machine, random forest, gradient boosting machine, and deep neural network-were trained and evaluated using a 5-fold cross-validation framework. Models were assessed for accuracy, sensitivity, specificity, F1-score, and area under the receiver operating characteristic curve. Shapley additive explanations values and activation maximization techniques were applied to interpret model predictions.

Results: The deep neural network model demonstrated superior performance with an accuracy of 89% (95% CI 86%-92%) and an area under the receiver operating characteristic curve of 0.95 (95% CI 0.93-0.97), outperforming traditional diagnostic methods by 15% (P<.001). Key predictive features included altered functional connectivity between the dorsolateral prefrontal cortex, anterior cingulate cortex, and limbic regions. The model achieved 78% sensitivity (95% CI 71%-85%) in identifying individuals who developed MDD within a 2-year follow-up period, demonstrating good generalizability across datasets.

Conclusions: Our findings highlight the potential of artificial intelligence-driven approaches for the early detection of MDD, with implications for improving early intervention strategies. While promising, these tools should complement rather than replace clinical expertise, with careful consideration of ethical implications such as patient privacy and model biases.

Background: Anxiety has become a significant health concern affecting mental and physical well-being, with state anxiety (s-anxiety)-a transient emotional response-linked to adverse cardiovascular and long-term health outcomes. Traditional physiological monitoring lacks the contextual sensitivity needed to assess anxiety in real time. Electrooculography (EOG) and electrodermal activity (EDA), 2 biosignals measurable by wearables, offer promising avenues for identifying biomarkers of s-anxiety in naturalistic environments.

Objective: This study aims to identify novel biomarkers of s-anxiety using EOG and EDA signals collected in real-world conditions. We further explore how noninvasive wearable technology can enable real-time monitoring of physiological responses during induced stress, focusing on distinguishing true anxiety-related signals from artifacts in noisy environments.

Methods: Our study presents two datasets: (1) the EOG signal blink identification dataset Blink Identification Electrooculography Dataset (BLINKEO), containing both true blink events and motion artifacts, and (2) the EOG and EDA signals dataset Emotion, Electrooculography, and Electrodermal Activity Monitoring in Cold Pressor Conditions Dataset (EMOCOLD), capturing physiological responses from a cold pressor test (CPT). From analyzing blink rate variability, skin conductance peaks, and associated arousal metrics, we identified multiple new anxiety-specific biomarkers. Shapley additive explanations (SHAP) were used to interpret and refine our model, enabling a robust understanding of the biomarkers that correlate strongly with s-anxiety.

Results: BLINKEO feature analysis achieved a classification accuracy of 98.17% and F1-score of 0.87 in distinguishing blinks from noise. In the EMOCOLD, survey results confirmed elevated anxiety and affectivity during the CPT, which normalized during recovery. SHAP analysis revealed that specific EDA features (eg, Hjorth activity and spectral entropy) and EOG features (eg, opening phase energy and signal height) consistently contributed to accurate predictions of s-anxiety and affectivity. Contextual combinations of features outperformed single-feature analyses, revealing relationships critical for robust biomarker identification.

Conclusions: These results suggest that a combined analysis of EOG and EDA data offers significant improvements in detecting real-time anxiety markers, underscoring the potential of wearables in personalized health monitoring and mental health intervention strategies. This work contributes to the development of context-sensitive models for anxiety assessment, promoting more effective applications of wearable technology in health care.

Background: Tuberculosis (TB) remains a significant global health challenge, as current diagnostic methods are often resource-intensive, time-consuming, and inaccessible in many high-burden communities, necessitating more efficient and accurate diagnostic methods to improve early detection and treatment outcomes.

Objective: This study aimed to evaluate the performance of 6 convolutional neural network architectures-Visual Geometry Group-16 (VGG16), VGG19, Residual Network-50 (ResNet50), ResNet101, ResNet152, and Inception-ResNet-V2-in classifying chest x-ray (CXR) images as either normal or TB-positive. The impact of data augmentation on model performance, training times, and parameter counts was also assessed.

Methods: The dataset of 4200 CXR images, comprising 700 labeled as TB-positive and 3500 as normal cases, was used to train and test the models. Evaluation metrics included accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve. The computational efficiency of each model was analyzed by comparing training times and parameter counts.

Results: VGG16 outperformed the other architectures, achieving an accuracy of 99.4%, precision of 97.9%, recall of 98.6%, F1-score of 98.3%, and area under the receiver operating characteristic curve of 98.25%. This superior performance is significant because it demonstrates that a simpler model can deliver exceptional diagnostic accuracy while requiring fewer computational resources. Surprisingly, data augmentation did not improve performance, suggesting that the original dataset's diversity was sufficient. Models with large numbers of parameters, such as ResNet152 and Inception-ResNet-V2, required longer training times without yielding proportionally better performance.

Conclusions: Simpler models like VGG16 offer a favorable balance between diagnostic accuracy and computational efficiency for TB detection in CXR images. These findings highlight the need to tailor model selection to task-specific requirements, providing valuable insights for future research and clinical implementations in medical image classification.

Background: Thailand's pharmaceutical industry is prioritizing innovation and self-reliance through the development of incrementally modified drugs (IMDs), particularly sustained-release dosage forms. However, the financial feasibility of IMD development remains underexplored.

Objective: This study evaluates the financial feasibility of developing sustained-release IMDs in Thailand, focusing on costs, timelines, and investment requirements to inform strategic decision-making.

Methods: A mixed methods approach was used, combining literature reviews, expert interviews, and financial modeling. Two scenarios were analyzed: (1) only development (phase I) and (2) full clinical trials (phase I to III). Sensitivity analysis was used to assess the impact of key variables on financial feasibility.

Results: The research and development (R&D) process for sustained-release IMDs takes 7 years for phase I-only development, costing US $1.46-3.09 million, and 11 years for full clinical trials, costing US $18.60-20.23 million. Process validation batches accounted for 60% of costs in phase I-only scenarios, while clinical trials represented 70% of costs in full clinical trial scenarios. The annual income required for a 5-year payback period ranged from US $0.20-1.80 million (phase I only) to US $3.01-27.11 million (full trials). Shorter R&D durations and longer payback periods substantially improved feasibility.

Conclusions: Developing sustained-release IMDs in Thailand involves substantial costs and extended timelines but offers a lower-risk alternative to new chemical entities. Strategic investments, efficient R&D processes, and supportive policies are essential to enhance feasibility and alignment with national goals of innovation and self-reliance.

Background: Academic bullying among junior doctors-characterized by repeated actions that undermine confidence, reputation, and career progression-is associated with adverse consequences for mental health and professional development.

Objective: This study aimed to investigate the prevalence and determinants of academic bullying among junior doctors in Sierra Leone.

Methods: We conducted a cross-sectional survey of 126 junior doctors at the University of Sierra Leone Teaching Hospitals Complex in Freetown between January 1 and March 30, 2024. Participants were selected through random sampling. Data were collected using a semistructured, self-administered questionnaire and analyzed with descriptive statistics and multivariable logistic regression.

Results: Of the 126 participants (n=77, 61.1% male; mean age 31.9, SD 5.05 years), 86 (68.3%) participants reported experiencing academic bullying. Among those, 55.8% (n=48) of participants experienced it occasionally and 36% (n=31) of participants experienced it very frequently. The most common forms were unfair criticism (n=63, 73.3%), verbal aggression (n=57, 66.3%), and derogatory remarks (n=41, 47.7%). Consultants and senior doctors were the main perpetrators, with incidents primarily occurring during ward rounds, clinical meetings, and academic seminars. No statistically significant predictors of bullying were found for gender (odds ratio 2.07, 95% CI 0.92-4.64; P=.08) or less than 2 years of practice (odds ratio 0.30, 95% CI 0.05-1.79; P=.19).

Conclusions: Academic bullying is widespread among junior doctors at the University of Sierra Leone Teaching Hospitals Complex. It has serious consequences for their mental health and professional development. There is an urgent need for clear and culturally appropriate policies, targeted training programs, confidential reporting systems, and leadership development. Promoting ethical leadership and fostering a culture of respect can help reduce incivility and burnout, leading to a healthier work environment for junior doctors.

Background: Access to contraception is a preventive measure against unplanned pregnancy and sexually transmitted infections; especially in sub-Saharan Africa where unmet need is a public health concern.

Objective: This study assessed the levels and predictors of knowledge, attitudes, and practices regarding contraception among female TV studies students in Nigeria.

Methods: This is a cross-sectional study conducted among female students of NTA TV College, Nigeria. Categorical sociodemographics, knowledge, attitude, and practice were presented as frequencies and proportions, while the continuous variables were presented as summary measures of central tendencies and dispersions. The primary outcome variable was the practices regarding contraception, while attitude and knowledge were secondary outcome variables, with sociodemographics as covariates. Predictors of good knowledge, attitude, and practice regarding contraception were determined by multivariable binary logistic regression, which was preceded by a bivariate regression analysis to determine candidate variables for the final model. A P value <.05 was determined to be statistically significant.

Results: There were 217 study participants with an average age of 22 (SD 2.6) years. Levels of good knowledge, attitude, and practice regarding contraception were reported in 55.3% (n=120), 47.5% (n=103), and 50.7% (n=110) of participants, respectively. The majority have had sex, used friends and the internet as their main sources of contraceptive information, and commonly used contraceptives such as condoms and oral contraceptive pills. The most common reason for not using contraceptives was fear of side effects or health risks. Being a young adult was a significant predictor (adjusted odds ratio [aOR] 2.6, 95% CI 1.0-6.7; P=.04) of good knowledge, while being a diploma student (aOR 2.4, 95% CI 1.2-4.6; P=.01), living off campus (aOR 2.1, 95% CI 1.0-4.4; P=.04), and good knowledge (aOR 3.8, 95% CI 2.1-6.9; P<.001) were significant predictors of good attitude. Being from the state's indigenous population (aOR 2.4, 95% CI 1.2-4.6; P=.01) and having engaged in sex (aOR 24.5, 95% CI 7.9-75.7; P<.001) were significant predictors of good contraception use.

Conclusions: Our study has shown relatively low levels of good knowledge, attitude, and practice regarding contraception and their predictors. Therefore, there is an urgent need to consistently improve advocacy, curricular development, and policies to improve knowledge, attitude, and practice regarding contraception and sexual and reproductive health services among young people.

Background: Rural health care providers face unique challenges such as limited specialist access and high patient volumes, making accurate diagnostic support tools essential. Large language models like GPT-3 have demonstrated potential in clinical decision support but remain understudied in pediatric differential diagnosis.

Objective: This study aims to evaluate the diagnostic accuracy and reliability of a fine-tuned GPT-3 model compared to board-certified pediatricians in rural health care settings.

Methods: This multicenter retrospective cohort study analyzed 500 pediatric encounters (ages 0-18 years; n=261, 52.2% female) from rural health care organizations in Central Louisiana between January 2020 and December 2021. The GPT-3 model (DaVinci version) was fine-tuned using the OpenAI application programming interface and trained on 350 encounters, with 150 reserved for testing. Five board-certified pediatricians (mean experience: 12, SD 5.8 years) provided reference standard diagnoses. Model performance was assessed using accuracy, sensitivity, specificity, and subgroup analyses.

Results: The GPT-3 model achieved an accuracy of 87.3% (131/150 cases), sensitivity of 85% (95% CI 82%-88%), and specificity of 90% (95% CI 87%-93%), comparable to pediatricians' accuracy of 91.3% (137/150 cases; P=.47). Performance was consistent across age groups (0-5 years: 54/62, 87%; 6-12 years: 47/53, 89%; 13-18 years: 30/35, 86%) and common complaints (fever: 36/39, 92%; abdominal pain: 20/23, 87%). For rare diagnoses (n=20), accuracy was slightly lower (16/20, 80%) but comparable to pediatricians (17/20, 85%; P=.62).

Conclusions: This study demonstrates that a fine-tuned GPT-3 model can provide diagnostic support comparable to pediatricians, particularly for common presentations, in rural health care. Further validation in diverse populations is necessary before clinical implementation.

Background: The increasing integration of artificial intelligence (AI) systems into critical societal sectors has created an urgent demand for robust privacy-preserving methods. Traditional approaches such as differential privacy and homomorphic encryption often struggle to maintain an effective balance between protecting sensitive information and preserving data utility for AI applications. This challenge has become particularly acute as organizations must comply with evolving AI governance frameworks while maintaining the effectiveness of their AI systems.

Objective: This paper aims to introduce and validate data obfuscation through latent space projection (LSP), a novel privacy-preserving technique designed to enhance AI governance and ensure responsible AI compliance. The primary goal is to develop a method that can effectively protect sensitive data while maintaining essential features necessary for AI model training and inference, thereby addressing the limitations of existing privacy-preserving approaches.

Methods: We developed LSP using a combination of advanced machine learning techniques, specifically leveraging autoencoder architectures and adversarial training. The method projects sensitive data into a lower-dimensional latent space, where it separates sensitive from nonsensitive information. This separation enables precise control over privacy-utility trade-offs. We validated LSP through comprehensive experiments on benchmark datasets and implemented 2 real-world case studies: a health care application focusing on cancer diagnosis and a financial services application analyzing fraud detection.

Results: LSP demonstrated superior performance across multiple evaluation metrics. In image classification tasks, the method achieved 98.7% accuracy while maintaining strong privacy protection, providing 97.3% effectiveness against sensitive attribute inference attacks. This performance significantly exceeded that of traditional anonymization and privacy-preserving methods. The real-world case studies further validated LSP's effectiveness, showing robust performance in both health care and financial applications. Additionally, LSP demonstrated strong alignment with global AI governance frameworks, including the General Data Protection Regulation, the California Consumer Privacy Act, and the Health Insurance Portability and Accountability Act.

Conclusions: LSP represents a significant advancement in privacy-preserving AI, offering a promising approach to developing AI systems that respect individual privacy while delivering valuable insights. By embedding privacy protection directly within the machine learning pipeline, LSP contributes to key principles of fairness, transparency, and accountability. Future research directions include developing theoretical privacy guarantees, exploring integration with federated learning systems, and e

Background: Pneumonia is a leading cause of mortality in children aged <5 years. While machine learning (ML) has been applied to pneumonia diagnostics, few studies have focused on predicting the need for escalation of care in pediatric cases. This study aims to develop an ML-based clinical decision support tool for predicting the need for escalation of care in community-acquired pneumonia cases.

Objective: The primary objective was to develop a robust predictive tool to help primary care physicians determine where and how a case should be managed.

Methods: Data from 437 children with community-acquired pneumonia, collected before the COVID-19 pandemic, were retrospectively analyzed. Pediatricians encoded key clinical features from unstructured medical records based on Integrated Management of Childhood Illness guidelines. After preprocessing with Synthetic Minority Oversampling Technique-Tomek to handle imbalanced data, feature selection was performed using Shapley additive explanations values. The model was optimized through hyperparameter tuning and ensembling. The primary outcome was the level of care severity, defined as the need for referral to a tertiary care unit for intensive care or respiratory support.

Results: A total of 437 cases were analyzed, and the optimized models predicted the need for transfer to a higher level of care with an accuracy of 77% to 88%, achieving an area under the receiver operator characteristic curve of 0.88 and an area under the precision-recall curve of 0.96. Shapley additive explanations value analysis identified hypoxia, respiratory distress, age, weight-for-age z score, and complaint duration as the most important clinical predictors independent of laboratory diagnostics.

Conclusions: This study demonstrates the feasibility of applying ML techniques to create a prognostic care decision tool for childhood pneumonia. It provides early identification of cases requiring escalation of care by combining foundational clinical skills with data science methods.