A 360° View for Large Language Models: Early Detection of Amblyopia in Children using Multi-View Eye Movement Recordings.

Amblyopia is a neurodevelopmental visual disorder that affects approximately 3-5% of children globally and it can lead to vision loss if it is not diagnosed and treated early. Traditional diagnostic methods, which rely on subjective assessments and expert interpretation of eye movement recordings presents challenges in resource-limited eye care centers. This study introduces a new approach that integrates the Gemini large language model (LLM) with eye-tracking data to develop a classification tool for diagnosis of patients with amblyopia. The study demonstrates: (1) LLMs can be successfully applied to the analysis of fixation eye movement data to diagnose patients with amblyopia; and (2) Input of medical subject matter expertise, introduced in this study in the form of medical expert augmented generation (MEAG), is an effective adaption of the generic retrieval augmented generation (RAG) approach for medical applications using LLMs. This study introduces a new multi-view prompting framework for ophthalmology applications that incorporates fine granularity feedback from pediatric ophthalmologist together with in-context learning to report an accuracy of 80% in diagnosing patients with amblyopia. In addition to the binary classification task, the classification tool is generalizable to specific subpopulations of amblyopic patients based on severity of amblyopia, type of amblyopia, and with or without nystagmus. The model reports an accuracy of: (1) 83% in classifying patients with moderate or severe amblyopia, (2) 81% in classifying patients with mild or treated amblyopia; and (3) 85% accuracy in classifying patients with nystagmus. To the best of our knowledge, this is the first study that defines a multi-view prompting framework with MEAG to analyze eye tracking data for the diagnosis of amblyopic patients.

Machine Learning Interpretability Methods to Characterize the Importance of Hematologic Biomarkers in Prognosticating Patients with Suspected Infection.

Early detection of sepsis in patients admitted to the emergency department (ED) is an important clinical objective as early identification and treatment can help reduce morbidity and mortality rate of 20% or higher. Hematologic changes during sepsis-associated organ dysfunction are well established and a new biomarker called Monocyte Distribution Width (MDW) has been recently approved by the US Food and Drug Administration for sepsis. However, MDW, which quantifies monocyte activation in sepsis patients, is not a routinely reported parameter and it requires specialized proprietary laboratory equipment. Further, the relative importance of MDW as compared to other routinely available hematologic parameters and vital signs has not been studied, which makes it difficult for resource constrained hospital systems to make informed decisions in this regard. To address this issue, we analyzed data from a cohort of ED patients (n=10,229) admitted to a large regional safety-net hospital in Cleveland, Ohio with suspected infection who later developed poor outcomes associated with sepsis. We developed a new analytical framework consisting of seven data models and an ensemble of high accuracy machine learning (ML) algorithms (accuracy values ranging from 0.83 to 0.90) for the prediction of outcomes more common in sepsis than uncomplicated infection (3-day intensive care unit stay or death). To characterize the contributions of individual hematologic parameters, we applied the Local Interpretable Model-Agnostic Explanation (LIME) and Shapley Additive Value (SHAP) interpretability methods to the high accuracy ML algorithms. The ML interpretability results were consistent in their findings that the value of MDW is grossly attenuated in the presence of other routinely reported hematologic parameters and vital signs data. Further, this study for the first time shows that complete blood count with differential (CBC-DIFF) together with vital signs data can be used as a substitute for MDW in high accuracy ML algorithms to screen for poor outcomes associated with sepsis.

MaTiLDA: An Integrated Machine Learning and Topological Data Analysis Platform for Brain Network Dynamics.

Topological data analysis (TDA) combined with machine learning (ML) algorithms is a powerful approach for investigating complex brain interaction patterns in neurological disorders such as epilepsy. However, the use of ML algorithms and TDA for analysis of aberrant brain interactions requires substantial domain knowledge in computing as well as pure mathematics. To lower the threshold for clinical and computational neuroscience researchers to effectively use ML algorithms together with TDA to study neurological disorders, we introduce an integrated web platform called MaTiLDA. MaTiLDA is the first tool that enables users to intuitively use TDA methods together with ML models to characterize interaction patterns derived from neurophysiological signal data such as electroencephalogram (EEG) recorded during routine clinical practice. MaTiLDA features support for TDA methods, such as persistent homology, that enable classification of signal data using ML models to provide insights into complex brain interaction patterns in neurological disorders. We demonstrate the practical use of MaTiLDA by analyzing high-resolution intracranial EEG from refractory epilepsy patients to characterize the distinct phases of seizure propagation to different brain regions. The MaTiLDA platform is available at: https://bmhinformatics.case.edu/nicworkflow/MaTiLDA.

Machine Learning Interpretability Methods to Characterize Brain Network Dynamics in Epilepsy.

The rapid adoption of machine learning (ML) algorithms in a wide range of biomedical applications has highlighted issues of trust and the lack of understanding regarding the results generated by ML algorithms. Recent studies have focused on developing interpretable ML models and establish guidelines for transparency and ethical use, ensuring the responsible integration of machine learning in healthcare. In this study, we demonstrate the effectiveness of ML interpretability methods to provide important insights into the dynamics of brain network interactions in epilepsy, a serious neurological disorder affecting more than 60 million persons worldwide. Using high-resolution intracranial electroencephalogram (EEG) recordings from a cohort of 16 patients, we developed high accuracy ML models to categorize these brain activity recordings into either seizure or non-seizure classes followed by a more complex task of delineating the different stages of seizure progression to different parts of the brain as a multi-class classification task. We applied three distinct types of interpretability methods to the high-accuracy ML models to gain an understanding of the relative contributions of different categories of brain interaction patterns, including multi-focii interactions, which play an important role in distinguishing between different states of the brain. The results of this study demonstrate for the first time that post-hoc interpretability methods enable us to understand why ML algorithms generate a given set of results and how variations in value of input values affect the accuracy of the ML algorithms. In particular, we show in this study that interpretability methods can be used to identify brain regions and interaction patterns that have a significant impact on seizure events. The results of this study highlight the importance of the integrated implementation of ML algorithms together with interpretability methods in aberrant brain network studies and the wider domain of biomedical research.

Utilization of adolescent friendly health services and its associated factors among higher secondary students in mid-western Himalayan mountainous district of Nepal.

Adolescent friendly health services (AFHS) are designed to make health services accommodate the unique needs of adolescents. AFHS are characterized by three basic characteristics (programmatic, health facilities and health service providers) that should be applied. However, limited is known about the use of AFHS in the context of Nepal. This study aimed to assess the extent of AFHS utilization and associated factors among higher secondary students in the Jumla district of Nepal. A cross-sectional quantitative study was conducted in October-November 2017. Data were collected from a random sample of 528 aged 16-19 years old using a self-administered survey in their classroom. Adjusted Odds Ratios (AOR) and a 95% confidence level were estimated to measure the strength of association between the outcome variable (utilization of AFHS) and independent variable using multivariable logistic regression. Knowledge related to AFHS, measured by a seven-item scale, was based on information about the availability of AFHS. More than two-thirds (67.05%) of adolescents had utilized AFHS at least once in the last twelve months before the survey. In multivariable logistic regression analysis, knowledge level [AOR = 14.796, 95%CI (5.326-41.099)], cost of services [AOR = 2.971, 95%CI (1.764-5.003)], satisfaction from services [AOR = 1.817, 95%CI (1.037-3.185)] and availability of waiting room [AOR = 1.897, 95%CI (1.096-3.283)] were significantly associated with the utilization of AFHS. The utilization of AFHS was less than the country's target of universal utilization in this study. Adolescents' knowledge level about AFHS was importantly associated with its utilization. Utilization increases with lower service costs, client satisfaction, and availability of waiting rooms in the health facility. The health planners should make efforts to create a conducive environment for the adolescent by training the AFHS providers, particularly those who work in government institutions, and strengthening the awareness creation strategies among adolescents to increase the utilization of the services.

Ontology-based feature engineering in machine learning workflows for heterogeneous epilepsy patient records.

Biomedical ontologies are widely used to harmonize heterogeneous data and integrate large volumes of clinical data from multiple sources. This study analyzed the utility of ontologies beyond their traditional roles, that is, in addressing a challenging and currently underserved field of feature engineering in machine learning workflows. Machine learning workflows are being increasingly used to analyze medical records with heterogeneous phenotypic, genotypic, and related medical terms to improve patient care. We performed a retrospective study using neuropathology reports from the German Neuropathology Reference Center for Epilepsy Surgery at Erlangen, Germany. This cohort included 312 patients who underwent epilepsy surgery and were labeled with one or more diagnoses, including dual pathology, hippocampal sclerosis, malformation of cortical dysplasia, tumor, encephalitis, and gliosis. We modeled the diagnosis terms together with their microscopy, immunohistochemistry, anatomy, etiologies, and imaging findings using the description logic-based Web Ontology Language (OWL) in the Epilepsy and Seizure Ontology (EpSO). Three tree-based machine learning models were used to classify the neuropathology reports into one or more diagnosis classes with and without ontology-based feature engineering. We used five-fold cross validation to avoid overfitting with a fixed number of repetitions while leaving out one subset of data for testing, and we used recall, balanced accuracy, and hamming loss as performance metrics for the multi-label classification task. The epilepsy ontology-based feature engineering approach improved the performance of all the three learning models with an improvement of 35.7%, 54.5%, and 33.3% in logistics regression, random forest, and gradient tree boosting models respectively. The run time performance of all three models improved significantly with ontology-based feature engineering with gradient tree boosting model showing a 93.8% reduction in the time required for training and testing of the model. Although, all three models showed an overall improved performance across the three-performance metrics using ontology-based feature engineering, the rate of improvement was not consistent across all input features. To analyze this variation in performance, we computed feature importance scores and found that microscopy had the highest importance score across the three models, followed by imaging, immunohistochemistry, and anatomy in a decreasing order of importance scores. This study showed that ontologies have an important role in feature engineering to make heterogeneous clinical data accessible to machine learning models and also improve the performance of machine learning models in multilabel multiclass classification tasks.

Prevalence of low birth weight and its associated factors: Hospital based cross sectional study in Nepal.

Low birth weight is still an important public health problem worldwide. It is a major contributor to neonatal death in developing countries, including Nepal. The government of Nepal has developed and implemented different programs to improve maternal and neonatal health, including baby's birth weight. However, low birth weight is a major maternal and child health challenge. Maternal factors determining the birth weight of neonates have been poorly assessed in previous studies in Nepal. Thus, this study aims to assess the prevalence and risk factors associated with low birth weight in Nepal. An institution-based descriptive cross-sectional study was carried out in Paropakar Maternity Hospital and Tribhuvan University Teaching Hospital of Kathmandu district among 308 postnatal mothers. The data was collected through the face-to-face interview technique. The data was entered in EpiData 3.1 and exported to Statistical Package and Service Solutions version 21 for analysis. Multivariate logistic regression was used to obtain an adjusted odds ratio, while p-value < 0.05 with 95% Confidence Interval (CI) was considered significant. The findings showed that 15.3% of the children had low birth weight. The mean and standard deviation of childbirth weight was 2.96±0.59 kg. Mothers belonged to Dalit ethnic (AOR = 2.9, 95% CI = 1.2-7.1), Antenatal Care visited three or fewer (AOR = 2.6, 95%CI = 1.0-6.6) and did not comply with Iron and Folic Acid supplementation (AOR = 2.1, 95% CI = 1.0-4.4) were significantly associated with low birth weight. Nearly one in every six children had low birth weight. Maternal health services such as antenatal care and compliance with a recommended dose of maternal micronutrients significantly impact on birth weight. Maternal and neonatal health programs should consider these factors to reduce adverse birth outcomes in Nepal.

Patient satisfaction and its socio-demographic correlates in a tertiary public hospital in Nepal: a cross-sectional study.

BACKGROUND: Patient satisfaction is one proxy indicator of the health care quality; however, enhancing patient satisfaction in low-income settings is very challenging due to the inadequacy of resources as well as low health literacy among patients. In this study, we assess patient satisfaction and its correlates in a tertiary public hospital in Nepal. METHODS: We conducted a cross sectional study at outpatient department of Bhaktapur Hospital of Nepal. To recruit participants for the study, we applied a systematic random sampling method. Our study used a validated Patient Satisfaction Questionnaire III (PSQ-III) developed by RAND Corporation including various contextual socio-demographic characteristics. We calculated mean score and percentages of satisfaction across seven dimensions of patient satisfaction. To determine the association between various dimensions of patient satisfaction and socio-demographic characteristics of the patient, we used a multi-ordinal logistic regression. RESULTS: Among 204 patients, we observed a wide variation in patient satisfaction across seven dimensions. About 39% of patients were satisfied in the dimension of general satisfaction, 92% in interpersonal manner, and 45% in accessibility and convenience. Sociodemographic factors such as age (AOR: 6.42; CI: 1.30-35.05), gender (AOR: 2.81; CI: 1.41-5.74), and ethnicity (AOR: 0.26; CI: 0.08-0.77) were associated with general satisfaction of the patients. Other sociodemographic variables such as education, occupation, and religion were associated with a majority of the dimensions of patient satisfaction (p < 0.05). Age was found to be the strongest predictor of patient satisfaction in five out of seven dimensions. CONCLUSIONS: We concluded that patient satisfaction varies across different dimensions. Therefore, targeted interventions that direct to improve the dimensions of patient satisfaction where the proportion of satisfaction is low are needed. Similar studies should be conducted regularly at different levels of health facilities across the country to capture a wider picture of patient satisfaction at various levels.

Frontline Healthcare Workers' Knowledge and Perception of COVID-19, and Willingness to Work during the Pandemic in Nepal.

This study investigated the contextual factors associated with the knowledge, perceptions, and the willingness of frontline healthcare workers (FHWs) to work during the COVID-19 pandemic in Nepal among a total of 1051 FHWs. Multivariable logistic regression analysis was applied to identify independent associations between predictors and outcome variables. Of the total study subjects, 17.2% reported inadequate knowledge on COVID-19, 63.6% reported that they perceived the government response as unsatisfactory, and 35.9% showed an unwillingness to work during the pandemic. Our analyses demonstrated that FHWs at local public health facilities, pharmacists, Ayurvedic health workers (HWs), and those with chronic diseases were less likely, and male FHWs were more likely, to have adequate knowledge of COVID-19. Likewise, nurses/midwives, public health workers, FHWs from Karnali and Far-West provinces, and those who had adequate knowledge of COVID-19 were more likely to have satisfactory perceptions towards the government response. Further, FHWs-paramedics, nurse/midwives, public health workers, laboratory workers-FHWs from Karnali Province and Far-West Province, and those with satisfactory perceptions of government responses to COVID-19 were predictors of willingness to work during the COVID-19 pandemic. These results suggest that prompt actions are required to improve FHWs' knowledge of COVID-19, address negative perceptions of government responses, and motivate them through specific measures to provide healthcare services during the pandemic.

Socio-demographic and Healthcare-seeking Predictors of Undernutrition among Children Under-five Years of Age in a Western District of Nepal.

BACKGROUND: Undernutrition is highly prevalent in Nepal, which interferes with physical and mental development among children. It is one of the severe health problems contributing to the significant portion of the disease burden. This study aimed to explore socio-demographic and healthcare-seeking related predictors of undernutrition among children under five years old in Dang, Nepal. METHODS: This was a descriptive cross-sectional study. A sample of 426 children was participated through stratified proportionate random sampling to identify socio-demographics and healthcare-seeking predictors of undernutrition. Multivariable regression was applied to identify the independent predictors of undernutrition. RESULTS: This study found that children below 24 months of age were more likely to be undernourished than children aged 24-36 months. Female children (OR=2.32, 95% CI: 1.19-4.54), illiterate or non-formally educated women (OR=4.09, 95% CI: 1.84-9.08), mother's occupation other than a housewife (OR=13.05, 95% CI: 4.19-40.68), labor work of father (OR=2.40, 95% CI: 1.04-5.57) had increased risk of undernutrition among children. Similarly, food insufficiency from their land, antenatal care visit, postnatal care visit, and delivery place were significantly associated with childhood undernutrition among children.  Conclusions: The study showed that undernutrition among children is associated with age and gender of children, educational attainment of the mother, food sufficiency, health-seeking practices of the mother during pregnancy, delivery, and postnatal. Socio-demographics and health-seeking practices related predictors must be explicitly considered to address undernutrition among children under the age of five years.

Sociodemographic correlates of antenatal care visits in Nepal: results from Nepal Demographic and Health Survey 2016.

BACKGROUND: Good quality antenatal care visits are crucial to reduce maternal mortality and improve overall maternal and neonatal health outcomes. A previous study on antenatal care visits analyzed the nationally representative data of 2011; however, no studies have been conducted recently in Nepal. Therefore, we analyzed the sociodemographic correlates of the frequency and quality of antenatal care among Nepalese women from the nationally representative data of 2016. METHODS: We analyzed data obtained from the Nepal Demography Health Survey (2016) on antenatal care for 2761 women who had one or more births in the past three years. Our study defined 'good quality antenatal care' as at least a 75% score on a composite metric which was obtained by adding the weighted scores assigned to the twelve recommended components of antenatal care. We analyzed the factors associated with the frequency and quality of antenatal care by using multiple Poisson regression and multiple logistic regression. RESULTS: While 70% of the Nepalese women surveyed had at least four antenatal care visits, only 21% of these women received good-quality antenatal care. We found that the educated women (APR: 1.12; CI: 1.05-1.19) and the women of rich wealth index (APR: 1.27; CI: 1.18-1.37) were more likely to receive a higher number of antenatal visits. In contrast, women living in rural areas (APR: 0.92; CI: 0.87-0.98), and those who had more than two children (APR: 0.88; CI: 0.83-0.93) were less likely to receive a higher number of antenatal visits. Regarding the quality of antenatal care, educated women (AOR: 1.51; CI: 1.09-2.08), women who had educated husbands (AOR: 2.11; CI: 1.38-3.22), women of rich wealth index (AOR: 1.58; CI: 1.13-2.20) and women who had intended pregnancy (APR: 1.69; CI: 1.23-2.34), were more likely to receive good-quality antenatal care. CONCLUSIONS: Observing a wide variation in the coverage of different components of antenatal care, concerned stakeholders could tailor the interventions by focusing on components with lower use. Because we found an association of myriad sociodemographic factors with the frequency and quality of antenatal care, targeted interventions are necessary.