Search

This study introduces the first-ever self-explanatory interface for diagnosing diabetes patients using machine learning. We propose four classification models (Decision Tree (DT), K-nearest Neighbor (KNN), Support Vector Classification (SVC), and Extreme Gradient Boosting (XGB)) based on the publicly available diabetes dataset. To elucidate the inner workings of these models, we employed the machine learning interpretation method known as Shapley Additive Explanations (SHAP). All the models exhibited commendable accuracy in diagnosing patients with diabetes, with the XGB model showing a slight edge over the others. Utilising SHAP, we delved into the XGB model, providing in-depth insights into the reasoning behind its predictions at a granular level. Subsequently, we integrated the XGB model and SHAP’s local explanations into an interface to predict diabetes in patients. This interface serves a critical role as it diagnoses patients and offers transparent explanations for the decisions made, providing users with a heightened awareness of their current health conditions. Given the high-stakes nature of the medical field, this developed interface can be further enhanced by including more extensive clinical data, ultimately aiding medical professionals in their decision-making processes.

Dengue is a common viral disease in tropical and subtropical countries. The clinical manifestation of dengue has a wide spectrum, from asymptomatic seroconversion to severe dengue infection. Severe dengue is defined as dengue with the presence of specific symptoms, including severe plasma leakage leading to shock or the accumulation of fluids with respiratory distress, severe bleeding, and severe organ impairment. Examining the progression of shock with the integration of patients’ physiological information and biochemical parameters would help in understanding the progression of the disease and early detection of shock. In this study, physiological patient data diagnosed with dengue are collected from a University Malaya Medical Centre’s electronic record. A prediction model learned from the measurement of a patient’s physiological data is the basis for effective treatment and prevention of shock development in critically ill patients. Hence, this study presents the predictive performance of machine learning algorithms to estimate the risk of shock development among dengue patients. Logistic regression, decision trees, support vector machines and neural networks are evaluated. Lastly, ensemble learnings of bagging and boosting are also applied to the weak learner to optimize performance. The experimental results show that the bagging algorithm outperforms other competing methods with a 14.5% improvement from the individual decision tree. The full blood count (FBC) specifically haemoglobin (Hb) on day 2 is found to be a strong predictor for severe dengue occurrence.

It is critical for healthcare providers to accurately determine lung cancer patients' prognostics and develop customized treatment plans. However, lung cancer has proven to be a complex disease, and every patient responds differently to treatment options, making survivability predictions highly challenging. This study proposes a holistic machine learning model that can assist healthcare providers in predicting the temporal effects of lung cancer-related factors on one-, five-, and ten-year survival rates. Variable selection algorithms such as genetic algorithm (GA) and Baruta are employed along with data balancing methods to achieve parsimonious models for survival prediction. Classification results are obtained through logistic regression and extreme gradient boosting algorithms followed by an information fusion technique to combine the classification results and identify the temporal effects of lung cancer variables over time. Results demonstrate that the prediction power of the classification models improved as the survival period increased. The models trained using the GA and intersection variable sets generated better average prediction scores. The study contributes to the cancer literature by analyzing the varying temporal impacts of lung cancer variables over varying time periods. Medical professionals can use these findings to understand better the longitudinal characteristics of lung cancer patients’ survival indicators.

This study addresses patient unpunctuality, a major concern affecting patient waiting time, resource utilization, and quality of care. We develop and compare four machine learning models, including multinomial logistic regression, decision tree, random forest, and artificial neural network, to accurately predict patient arrival patterns and aid efficient scheduling. These models are analyzed using the explainable artificial intelligence approach and the Shapley additive explanations model, promoting comprehension and trust in our algorithmic results. Using three years of appointment data from a psychiatric clinic, we identify the travel distance, appointment lead time, patient’s age, Body Mass Index (BMI), and certain mental diagnoses as significant factors affecting the patient’s unpunctuality. Despite the good predictive potential of machine learning algorithms, no single model excels in all performance metrics. The study proposes implementing these machine learning techniques and the explainable artificial intelligence tool into the clinic’s appointment system as a decision support system to minimize patient unpunctuality.

Artificial Intelligence (AI) solutions have been widely used in healthcare, and recent developments in deep neural networks have contributed to significant advances in medical image processing. Much ongoing research is aimed at helping medical practitioners by providing automated systems to analyze images and diagnose acute diseases, such as brain tumors, bone cancer, breast cancer, bone fracture, and many others. This comprehensive review delivers an overview of recent advances in medical imaging using deep neural networks. In addition to the comprehensive literature review, a summary of openly available data sources and future research directions are outlined.

Rheumatoid arthritis (RA), one of the most common chronic inflammatory diseases, is an autoimmune disease that causes cartilage, bone damage, and disability. RA has become a concern, affecting more than 1.5% of the global population. Immunomodulatory or anti-inflammatory targets are the lead candidates in existing clinical trials to develop RA drugs. These targets were reported to be ineffective in reducing symptoms, slowing disease progression, and causing adverse side effects. Therefore, this study focuses on addressing the continuing need for safer, more selective, and more effective treatment options for RA by identifying novel targets and therapeutic approaches. The study involves six important steps: retrieving a dataset from the Gene Expression Omnibus (GEO) database, file conversion, gene differentiation (up-regulated and downregulated), Entrez gene ID retrieval for Search Tool for the Retrieval of Interacting Genes (STRING), protein–protein interaction (PPI) network delineation via STRING, drugs identification via Drug Bank, and relationship establishment between networks and drugs via The Drug–Gene​ Interaction Database (DGIdb). Nine protein hubs, three up-regulated and six down-regulated were identified as interacting with 8 and 18 drugs, respectively. Seven protein sites have never been directly associated with RA and could serve as potential therapeutic targets. Eleven drugs were shown to interact with these seven selected protein hubs as potential therapeutics for RA. These new targets and therapies will help researchers improve the RA treatment pipeline by adding these into the pipeline and bringing successful treatment options for RA patients.

Cardiovascular disease (CVD) is a common disorder frequently resulting in death. An increase in the death rate among adults is attributed to several factors, including smoking, high blood pressure, obesity, and cholesterol. Early diagnosis of CVDs can lower mortality rates. Algorithms that use machine learning and data mining offer the potential for finding risk variables and predicting CVD. Developing countries often need more CVD experts, and a high percentage of misdiagnosis. These concerns could be alleviated using an accurate and effective early-stage heart disease prediction system. This study explores the effectiveness of machine learning classifiers for diagnosing and detecting CVD. Several supervised machine-learning algorithms are investigated, and their performance and accuracy are compared. The Gradient Boosted Decision Tree (GBDT) with Binary Spotted Hyena Optimizer (BSHO) suggested in this work was used to rank and classify all attributes. Discrete optimization problems can be resolved using the binary form of SHO. The recommended method compresses the continuous location using a hyperbolic tangent function. The updated spotted hyena positions on the relevance score are utilized to find those with high heart disease predictions. The efficiency of the suggested model is then confirmed using the UCI dataset. The proposed GBDT-BSHO approach, with an accuracy of 97.89%, was significantly more effective than the comparative methods.

While previous studies have shown the potential value of predictive modelling for emergency care, few models have been implemented for producing near real-time predictions across various demand, utilisation and performance metrics. In this study, 33 independent Random Forest (RF) algorithms were developed to forecast 11 urgent care metrics over a 24-hour period across three hospital sites in an Integrated Care System (ICS) in South West England. Metrics included: ambulance handover delay; emergency department occupancy; and patients awaiting admission. Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and Symmetric Mean Absolute Percentage Error (SMAPE) were used to assess the performance of RF and compare it to two alternative models: naïve baseline (NB) and Auto-Regressive Integrated Moving Average (ARIMA). Using these measures, RF outperformed NB and ARIMA in 76% (N = 25/33) of urgent care metrics according to SMAPE, 88% (N = 29/33) according to MAE and 91% (N = 30/33) according to RMSE. The RFs developed in this study have been implemented within the local ICS, providing predictions on an hourly basis that can be accessed by local healthcare planners and managers.

Patient value in hospital care has become increasingly important over the last decade. This paper argues that patient value could be ameliorated by investing in patient agility. Patient agility constitutes the capabilities that enable hospitals to sense the health service needs of their patients and respond to the changing health service demands of these patients. This paper further posits that hospitals’ current sensing and responding capabilities are inherently reactive as they detect, act upon and pursue occurrences and patterns in the needs and demands of patients after these have already manifested themselves. Hence, we assert that this reactive nature thwarts hospitals in preventing negative trends or capitalizing on positive trends necessitating a shift to sensing and responding capabilities that are more predictive. We postulate that this necessary organizational shift can be facilitated by implementing and utilizing the power of predictive analytics and that this particular issue has not yet been adequately addressed by existing scientific literature. Based on this argument, an initial research proposal is established centering around the role of predictive analytics in fostering patient agility and patient value in hospitals. The initial research proposal sets out to examine this topic in an inclusive, integrated and comprehensive manner as it outlines a coherent conceptual model, incorporates a multi-stakeholder perspective and integrates biomedical as well as health service needs and demands. The initial proposal is meant to serve as a conceptual framework for future research as well as an initial point of departure for further refinement, improvement, elaboration, and discussion.

Brain cancer is one of the most deadly cancers, with a very low survival rate. By understanding the factors that lead to cancer spreading, practitioners can concentrate their efforts on providing the most effective treatment, and they can modify the treatment plan as necessary. Also, knowing the likelihood of a patient’s survival over a specified time period can enable them to make informed decisions about adjusting their routines, future investments, and other health-related decisions. The use of data-driven models in cancer research has gained increased popularity over the past several decades. Moreover, there is still much uncertainty surrounding the factors that contribute to survival of cancer, making it difficult to develop a model. The existing literature on brain cancer contains a variety of machine learning models. However, many of them lack a high degree of accuracy, and, in medical research, accuracy is essential to the proper guidance of treatment decisions. Therefore, we have proposed a framework comprising multiple phases of classical statistics and machine learning methods to find a parsimonious model with a high degree of accuracy (98.9%) for predicting brain cancer survivability. Furthermore, we develop a prototype web-based interactive tool to facilitate the practical implementation of the proposed model and provide a deeper understanding of how a particular factor affects survival when other factors remain unchanged. By integrating this tool into healthcare settings, medical professionals can rapidly detect potentially vulnerable patients, and it can also be useful in determining the most effective treatment plan.

Heart disease remains the leading cause of death, such that nearly one-third of all deaths worldwide are estimated to be caused by heart-related conditions. Advancing applications of classification-based machine learning to medicine facilitates earlier detection. In this study, the Classification and Regression Tree (CART) algorithm, a supervised machine learning method, has been employed to predict heart disease and extract decision rules in clarifying relationships between input and output variables. In addition, the study’s findings rank the features influencing heart disease based on importance. When considering all performance parameters, the 87% accuracy of the prediction validates the model’s reliability. On the other hand, extracted decision rules reported in the study can simplify the use of clinical purposes without needing additional knowledge. Overall, the proposed algorithm can support not only healthcare professionals but patients who are subjected to cost and time constraints in the diagnosis and treatment processes of heart disease.

Gathering diabetes-related data requires effort from the patient side to log specific events throughout experiments. This is an error-prone task that the patients usually handle by following a prescribed protocol. However, patients often do not follow the protocol, causing missing or imperfect data. This study investigates the possibility of generating Markov models from existing/logged data and using them for imputation. The models are used to infer information related to missing events (types/activities) in data recorded by diabetes patients, allowing for improvements in the quality and continuity of such data. Our results indicate that such an approach can help improve the quality of the data collected.

The paper focuses on the construction of an artificial intelligence-based heart disease detection system using machine learning algorithms. We show how machine learning can help predict whether a person will develop heart disease. In this paper, a python-based application is developed for healthcare research as it is more reliable and helps track and establish different types of health monitoring applications. We present data processing that entails working with categorical variables and conversion of categorical columns. We describe the main phases of application developments: collecting databases, performing logistic regression, and evaluating the dataset’s attributes. A random forest classifier algorithm is developed to identify heart diseases with higher accuracy. Data analysis is needed for this application, which is considered significant according to its approximately 83% accuracy rate over training data. We then discuss the random forest classifier algorithm, including the experiments and the results, which provide better accuracies for research diagnoses. We conclude the paper with objectives, limitations and research contributions.

Reviewing patients who return to the emergency department (ED) within 72 h (i.e., bounce-back) is a standard quality assurance procedure used to identify correctable system- and clinician-level causes for earlier-than-expected return to the ED and ultimately ensure patients’ safety. This study proposes a deep learning (DL) framework to automatically extract features from structured and unstructured Electronic Health Records (EHR) data of ED visits and predict patients who are likely to bounce-back. Data from 120,000+ visits to the ED of four major hospitals in New York city over 4+ years are used to validate the proposed framework. The DL model was able to predict 74.8% of the bounce-backs (AUROC= 0.766). Notably, our results show that leveraging state-of-the-art DL techniques to extract features from unstructured ED physician narrative notes and incorporating them (along with structured vitals and demographics data) in the prediction task can remarkably improve the outcome. In addition, a sensitivity analysis performed on the structured features of the model revealed that factors such as patients’ age, type of medical insurance, chronic conditions (especially renal, heart, and respiratory), heart failure, and head or neck trauma are among the top factors increasing the chances of patients’ unplanned return to the ED. The proposed framework may be used as a decision support tool to assist emergency care clinicians in the early identification of the high-risk patients who are likely to bounce-back and provide them with timely, appropriate care.

The negative impact of stroke in society has led to concerted efforts to improve the management and diagnosis of stroke. With an increased synergy between technology and medical diagnosis, caregivers create opportunities for better patient management by systematically mining and archiving the patients’ medical records. Therefore, it is vital to study the interdependency of these risk factors in patients’ health records and understand their relative contribution to stroke prediction. This paper systematically analyzes the various factors in electronic health records for effective stroke prediction. Using various statistical techniques and principal component analysis, we identify the most important factors for stroke prediction. We conclude that age, heart disease, average glucose level, and hypertension are the most important factors for detecting stroke in patients. Furthermore, a perceptron neural network using these four attributes provides the highest accuracy rate and lowest miss rate compared to using all available input features and other benchmarking algorithms. As the dataset is highly imbalanced concerning the occurrence of stroke, we report our results on a balanced dataset created via sub-sampling techniques.

Surgical management has a direct impact on the quality treatment of patients and the economic efficiency of hospitals. Surgeries are managed in hospitals according to the duration of surgery reported by the surgeons. Surgical management efficiency would improve if the duration could be predicted accurately. This study focuses on predicting the individual duration in surgical management. The data on 9567 surgical cases from the National Cancer Center Hospital Japan East illustrates hospital characteristics and forecasts the duration requirements using a machine learning algorithm. We obtain an adjusted coefficient of determination exceeding 0.7 and show the relationship between the hospital characteristics and duration in a case-mix classification framework. We also discuss the relationship between data characteristics and the prediction accuracy of machine learning. This study supports operating room managers when they predict surgical duration by machine learning in ascertaining data characteristics, understanding the prediction accuracy of machine learning, and predicting surgical duration more accurately.

Diabetes mellitus is often a hyperglycemic condition that poses a substantial threat to human health. Early diabetes detection decreases morbidity and mortality. Due to the scarcity of labeled data and the presence of oddities in diabetes datasets, it is exceedingly difficult to develop a trustworthy and accurate diabetes prognosis. The dataset and groupings of the features using the elbow and silhouette methods have been clustered using K-means. Various machine learning approaches have also been applied to the cluster-based dataset to predict diabetes. We propose an unsupervised cluster-based feature grouping model for early diabetes identification using an open-source dataset containing the data of 520 diabetic patients. On the cluster-based dataset and the complete dataset, the maximum Accuracy (ACC) is 99.57% and 99.03%, respectively. The best Precision, Recall, minimum mean squared error (MSE), maximum mean squared error (MSE), and F1-Score of 1.000 are obtained from multi-layer perceptron (MLP), random forest (RF), and k-Nearest Neighbors (KNN), 0.984 from random forest (RF) and support vector machine (SVM), 0.010 from RF, 0.067 from KNN, and 99.20% from RF, respectively. A comparison table displays the anticipated outcomes and highlights the aspects of this research that are most likely to occur as intended. The preprocessed data and codes are available on the GitHub repository to https://github.com/mhashiq/Early-stage-diabetes-risk-prediction.

Rural hospitals continue to face significant challenges, including declining inpatient admissions, which may have a deteriorating impact on rural hospitals’ financial viability and the health care of rural residents. We use descriptive and predictive analytics to analyze trends in the average daily census (ADC), a measure of inpatient volume, and identify factors that may be associated with the ADC in Pennsylvania’s rural hospitals. We use retrospective longitudinal data on Pennsylvania’s rural acute care hospitals from 2000 to 2019. Overall, Pennsylvania’s rural hospitals experienced a decline in ADC. Readmission index, registered nurse (RN) staffing, population over 65 years of age, and population per square mile were statistically significant and positively associated with ADC. The mortality index was statistically significant and negatively associated with ADC, and the payer mix did not seem to impact ADC. In addition, critical access hospitals (CAHs) had a statistically significant and negative association with ADC. The continuing decline in ADC may deteriorate rural hospitals’ financial stability and overall health in rural communities. Hospital administrators should consider increasing nurse staffing to improve the quality of care and patient communication to build trust. Policymakers should encourage more hospitals, especially the CAHs and smaller hospitals, to participate in the Pennsylvania Rural Health Model to ensure their financial stability.

Following the outbreak of the coronavirus epidemic in early 2020, municipalities, regional governments and policymakers worldwide had to plan their Non-Pharmaceutical Interventions (NPIs) amidst a scenario of great uncertainty. At this early stage of an epidemic, where no vaccine or medical treatment is in sight, algorithmic prediction can become a powerful tool to inform local policymaking. However, when we replicated one prominent epidemiological model to inform health authorities in a region in the south of Brazil, we found that this model relied too heavily on manually predetermined covariates and was too reactive to changes in data trends. Our four proposed models access data of both daily reported deaths and infections as well as take into account missing data (e.g., the under-reporting of cases) more explicitly, with two of the proposed versions also attempting to model the delay in test reporting. We simulated weekly forecasting of deaths from the period from 31/05/2020 until 31/01/2021, with first week data being used as a cold-start to the algorithm, after which we use a lighter variant of the model for faster forecasting. Because our models are significantly more proactive in identifying trend changes, this has improved forecasting, especially in long-range predictions and after the peak of an infection wave, as they were quicker to adapt to scenarios after these peaks in reported deaths. Assuming reported cases were under-reported greatly benefited the model in its stability, and modelling retroactively-added data (due to the “hot” nature of the data used) had a negligible impact on performance.

Major depressive disorder (MDD), also known as depression, is a common and serious mental disorder. It is characterized by a high rate of relapse or recurrence where a person might experience depressive episodes after being depression-free. Although numerous studies are proposed in the literature for depression recognition using video, to the best of our knowledge, only one preliminary study has been proposed in the literature for the automatic identification of signs of depression relapse using audiovisual cues without user intervention. In this paper, we propose a proof of concept of a deep learning-based approach for depression recognition and depression relapse prediction using videos of clinical interviews. We propose a correlation-based anomaly detection framework and a measure of similarity to depression where depression relapse is detected when the deep audiovisual patterns of a depression-free subject become close to the deep audiovisual patterns of depressed subjects. Thus, the correlation between the audiovisual encoding of a test subject and a deep audiovisual representation of depression is computed and is used for monitoring depressed subjects and for predicting relapse after depression. Very promising results are achieved with an accuracy of 80.99% and 82.55% respectively for relapse depression prediction on the DAIC-Woz dataset.

This study examines how asthma and other chronic obstructive lung diseases vary in different regions. Information pertaining influence of region as a proxy for the environment on asthma is not well established, yet important for healthcare, travel, and migration decisions. The study was based on data collected from the Oman Ministry of Health’s annual reports from 2010 to 2019. Incidence rates for eleven regions in the Sultanate of Oman were analyzed using statistical tools, including; analysis of variance and the logistic regression model to determine the effect of region on asthma and other chronic obstructive lung diseases. Overall, the incident rates differed significantly by region, implying a potential environmental difference. There was no significant variation by year. Over the ten years, the incidence rates stagnated between 250 and 300 per 10000 of the population but showed a decreasing trend between 2016 and 2019. The regression model shows that all other regions had significantly increased odd ratios compared to the Muscat region. In addition, five asthma-region classifications were identified from our analysis. Therefore, there is significant evidence of regional variation in the incidence of asthma and other chronic obstructive lung diseases. This confirmation of regional differences could imply differences in environmental characteristics. This study’s findings may guide decision-making toward managing asthma and chronic obstructive lung diseases.