Expert Systems with Applications

The purpose of this paper is to present a detailed investigation of the advantages of employing GraphLIME (Local Interpretable Model Explanations for Graph Neural Networks) for the trustworthy prediction of diabetes mellitus. Our pursuit involves identifying the strengths of GraphLIME combined with the attention-mechanism over the standard coupling of deep learning neural networks with the original LIME method. The system build this way, provided us a proficient method for extracting the most relevant features and applying the attention mechanism exclusively to those features. We have closely monitored the performance metrics of the two approaches and conducted a comparative analysis. Leveraging attention mechanisms, we have achieved an accuracy of 92.6% for the addressed problem. The models performance is meticulously demonstrated throughout the study, and the results are furthermore evaluated using the Receiver Operating Characteristic (ROC) curve. By implementing this technique on a dataset of 768 patients diagnosed with or without diabetes mellitus, we have successfully boosted the models performance by over 18%.

We present a digital twin framework for real-time glucose monitoring and forecasting in septic patients in intensive care units (ICUs). The framework combines advanced machine learning models trained on continuous glucose measurements with a dynamic transfer-learning workflow that enables rapid deployment to individual patients and supports personalized, adaptive, and predictive clinical decision-making. Built on a foundation model--a pretrained time-series transformer--the digital twin continuously updates its parameters as new patient data arrive and produces rolling near-term forecasts in real time. To assess adaptability and computational efficiency, we deployed the pretrained model to ten septic patients and evaluated multiple retraining strategies, including zero-shot inference, linear probing, and full and staged fine-tuning. Results show that the model can be initialized and personalized for a new patient within seconds on a standard laptop while achieving accurate glucose forecasts under varying data conditions. These findings demonstrate the feasibility of real-time model personalization in resource-constrained, high-acuity environments and highlight the potential of digital twins as scalable, AI-enabled platforms for continuous physiological monitoring, clinical decision support, and individualized treatment design in the ICU.

The demand for Continuous Glucose Monitoring systems is increasing among type 1 diabetic patients. Some companies are trying to improve the monitoring and the usability of these systems. One example is Abbott FreeStyle Libre, which provides a new concept of glucose monitoring called Flash Glucose Monitoring which is more affordable and does not need calibration. The increasing demand for these devices means an opportunity for data and computer scientists, who can contribute to the development of decision-making support systems based on the data collected from the devices. Type 1 diabetic patients that use FreeStyle Libre may enter the number of insulin and carbohydrates units that they are going to take before a meal. Using both the entered data and the blood glucose values collected by the device automatically, the application presented in this paper generates a report of the patients glucose patterns. In addition, it provides a web application that allows the user to upload the data obtained from the device and download the report on his computer or smartphone. The application uses decision trees to detect the patterns and entails a starting point in the creation of ensemble models with more predictive power, also based on decision trees. Furthermore, the methodology makes a segmentation of the data set in blocks, determined by the different meals done throughout the day, adding more information to the set of variables used to train the model. As a result, the application can discover repetitive patterns in the daily life of the patient, which can help to take early preventive measures for risk situations in a period close to the next meal.

This paper explores the use of machine learning (ML) for predictive modeling on clinical data from the EMory BrEast imaging Dataset (EMBED) [1] with a focus on ML model fairness analysis. The aim of this study is to develop and evaluate fair machine learning models that can accurately predict breast cancer risk. We trained and tested various machine-learning models. Our findings show that machine learning can be effective for predicting breast cancer risk or diagnosing breast cancer, and that fairness considerations are crucial in the development of such models. Overall, our study highlights the potential of machine learning for clinical applications while emphasizing the need for ethical and fair practices in this field.

BackgroundEarly diabetes detection remains challenging, requiring robust machine learning approaches that balance accuracy with clinical interpretability for effective diagnostic support. MethodsWe are proposing a novel Additive and Multiplicative Neurons Network (AMNN) that combines both additive and multiplicative computational pathways to capture complex nonlinear relationships in diabetes prediction. Using the PIMA Indians Diabetes dataset (n=768), we compared AMNN against nine established algorithms including XGBoost, KAN, and traditional neural networks. Data preprocessing included SMOTE oversampling for class imbalance, and model interpretability was enhanced through SHAP and LIME explainable AI techniques. ResultsThe AMNN model outperformed all baseline approaches, achieving 75.76% accuracy, a 76.18% F1-score, and an AUC-ROC of 0.8206. Across both traditional feature selection techniques and explainable AI analyses, glucose levels, BMI, age, and pregnancy count consistently emerged as the most influential predictors. ConclusionsThe AMNN framework demonstrates strong potential for diabetes prediction by balancing accuracy with clinical interpretability. The key predictors it highlights align closely with established medical knowledge, reinforcing confidence in its outputs and suitability for use in clinical decision-making workflows. This hybrid neural network approach represents a promising step toward transparent, AI-assisted diagnostic tools that can support healthcare professionals in practice.

BackgroundPatients with type 2 diabetes mellitus (T2DM) who have severe hypoglycemia (SH) poses a considerable risk of long-term death, demanding urgent medical attention. Accurate prediction of SH remains challenging due to its multifactorial nature, contributed from factors such as medications, lifestyle choices, and metabolic measurements. MethodIn this study, we propose a systematic approach to improve the robustness and accuracy of SH predictions using machine learning models, guided by clinical feature selection. Our focus is on developing one-year SH prediction models using both semi-supervised learning and supervised learning algorithms. Utilizing the clinical trial, namely Action to Control Cardiovascular Risk in Diabetes, which involves electronic health records for over 10,000 individuals, we specifically investigate adults with T2DM who are at an increased risk of cardiovascular complications. ResultsOur results indicate that the application of a multi-view co-training method, incorporating the random forest algorithm, improves the specificity of SH prediction, while the same setup with Naive Bayes replacing random forest demonstrates better sensitivity. Our framework also provides interpretability of machine learning (XAI) models by identifying key predictors for hypoglycemia, including fast plasma glucose, hemoglobin A1c, general diabetes education, and NPH or L insulins. ConclusionBy enhancing prediction accuracy and identifying crucial predictive features, our study contributes to advancing the understanding and management of hypoglycemia in this population.

Disease dynamics, human mobility, and public policies co-evolve during a pandemic such as COVID-19. Understanding dynamic human mobility changes and spatial interaction patterns are crucial for understanding and forecasting COVID-19 dynamics. We introduce a novel graph-based neural network(GNN) to incorporate global aggregated mobility flows for a better understanding of the impact of human mobility on COVID-19 dynamics as well as better forecasting of disease dynamics. We propose a recurrent message passing graph neural network that embeds spatio-temporal disease dynamics and human mobility dynamics for daily state-level new confirmed cases forecasting. This work represents one of the early papers on the use of GNNs to forecast COVID-19 incidence dynamics and our methods are competitive to existing methods. We show that the spatial and temporal dynamic mobility graph leveraged by the graph neural network enables better long-term forecasting performance compared to baselines.

Microarray cancer datasets are characterized by a large number of irrelevant, redundant, and noisy features, which can severely hinder the accuracy and efficiency of classification algorithms. Feature selection, as a crucial branch of feature engineering, aims to enhance classification performance by identifying and retaining only the most informative features. However, feature selection is an NP-hard problem, where conventional search strategies are often prone to premature convergence and local optima, resulting in increased computational burden. To address these challenges, global metaheuristic algorithms have been widely explored. The recently proposed Lion Optimization (LO) algorithm has shown promising results for continuous optimization problems, yet its design is not inherently suited for discrete feature selection tasks. To overcome this limitation, a binary variant of the LO algorithm, termed Binary Lion Optimization (BLO), is introduced for wrapper-based feature selection in microarray cancer data analysis. In this work, the Minimum Redundancy Maximum Relevance (mRMR) criterion is first employed as a filter method to identify an initial subset of relevant features, thereby reducing search complexity. The refined feature subset is then optimized using the BLO algorithm to achieve improved classification outcomes. The proposed mRMR-BLO framework was evaluated on several widely recognized cancer microarray datasets and benchmarked against four state-of-the-art binary optimization algorithms. Experimental results demonstrate that mRMR-BLO consistently identifies smaller yet highly discriminative feature subsets, while achieving competitive or superior prediction accuracy. These findings highlight the potential of mRMR-BLO as an effective and robust tool for high-dimensional microarray cancer classification.

Globally the devastating consequence of COVID-19 or Severe Acute Respiratory Syndrome-Coronavirus (SARS-CoV-2) has posed danger on the life of living beings. Doctors and scientists throughout the world are working day and night to combat the proliferation or transmission of this deadly disease in terms of technology, finances, data repositories, protective equipment, and many other services. Rapid and efficient detection of COVID-19 reduces the rate of spreading this deadly disease and early treatment improve the recovery rate. In this paper, we proposed a new framework to exploit powerful features extracted from the autoencoder and Gray Level Co-occurence Matrix (GLCM), combined with random forest algorithm for the efficient and fast detection of COVID-19 using computed tomographic images. The models performance is evident from its 97.78% accuracy, 96.78% recall, and 98.77% specificity.

Coronavirus is a global emergency as of May 2021. If not acted upon by drugs at the right time, coronavirus may result in the death of individuals. Hence early diagnosis is very important along the progress of the disease. This paper focuses on coronavirus detection using x-ray images, for automating the diagnosis pipeline using convolutional neural networks and transfer learning. This could be deployed in places where radiologists are not easily available in order to detect the disease at very early stages. In this study we propose our deep learning architecture for the classification task, which is trained with modified images, through multiple steps of preprocessing. Our classification method uses convolutional neural networks and transfer learning architecture for classifying the images. Our findings yield an accuracy value of 91.03%, precision of 89.76 %, recall value of 96.67% and F1 score of 93.09%.

Cervical cancer represents a pressing global health challenge, emphasizing the critical need for accurate and timely diagnostic methods to facilitate effective treatment and improve survival rates. In response to this challenge, the study presents CerViX-Net, an innovative classification framework designed to advance cervical cancer detection through enhanced computational efficiency and diagnostic accuracy. The development of CerViX-Net is motivated by the limitations of traditional diagnostic models, particularly in handling the computational and memory demands of large-scale data, while ensuring precise feature extraction and classification. CerViX-Net employs a hybrid deep learning architecture that combines the capabilities of ResNet50, EfficientNet-B0, and a Modified Vision Transformer (ViT) module. The ResNet50 branch extracts hierarchical features through stacked convolutional and identity blocks. In another path, the modified ViT module transforms image patches via linear projection, augments them with positional and class embeddings, and processes them using Parallel Transformer Encoder layers to model contextual relationships. Concurrently, EfficientNet-B0 utilizes MBConv blocks to extract multi-scale representations. The feature outputs from all three branches are integrated and passed through a classification head consisting of dropout layers and dense layers to ensure robust and accurate predictions. The proposed framework is rigorously evaluated on the Mendeley LBC dataset, achieving exceptional performance metrics with an accuracy of 99.69%, precision of 99.28%, recall of 99.48%, and an F1-score of 99.52%. The robustness of CerViX-Net is further validated on the SIPaKMeD and Herlev Pap Smear datasets, where it demonstrates comparable excellence, underscoring its efficacy and adaptability across diverse cytology datasets. Statistical validation using Friedman's test further reinforces its superiority over competing methods.

Tuberculosis is an infectious disease that is leadingto the death of millions of people across the world. The mortalityrate of this disease is high in patients suffering from immuno-compromised disorders. The early diagnosis of this disease cansave lives and can avoid further complications. But the diagnosisof TB is a very complex task. The standard diagnostic tests stillrely on traditional procedures developed in the last century. Theseprocedures are slow and expensive. So this paper presents anautomatic approach for the diagnosis of TB from posteroanteriorchest x-rays. This is a two-step approach, where in the first stepthe lung regions are segmented from the chest x-rays using thegraph cut method, and then in the second step the transfer learn-ing of VGG16 combined with Bi-directional LSTM is used forextracting high-level discriminative features from the segmentedlung regions and then classification is performed using a fullyconnected layer. The proposed model is evaluated using data fromtwo publicly available databases namely Montgomery Countryset and Schezien set. The proposed model achieved accuracy andsensitivity of 97.76%, 97.01%and 96.42%, 94.11%on Schezienand Montgomery county datasets. This model enhanced thediagnostic accuracy of TB by 0.7%and 11.68%on Schezien andMontgomery county datasets.

In clinical applications, neural networks must focus on and highlight the most important parts of an input image. Soft-Attention mechanism enables a neural network to achieve this goal. This paper investigates the effectiveness of Soft-Attention in deep neural architectures. The central aim of Soft-Attention is to boost the value of important features and suppress the noise-inducing features. We compare the performance of VGG, ResNet, Inception ResNet v2 and DenseNet architectures with and without the Soft-Attention mechanism, while classifying skin lesions. The original network when coupled with Soft-Attention outperforms the baseline[16] by 4.7% while achieving a precision of 93.7% on HAM10000 dataset [25]. Additionally, Soft-Attention coupling improves the sensitivity score by 3.8% compared to baseline[31] and achieves 91.6% on ISIC-2017 dataset [2]. The code is publicly available at github1.

The novel coronavirus disease (COVID-19) and pandemic has taken the world by surprise and simultaneously challenged the health infrastructure of every country. Governments have resorted to draconian measures to contain the spread of the disease despite its devastating effect on their economies and education. Tracking the novel coronavirus 2019 disease remains vital as it influences the executive decisions needed to tighten or ease restrictions meant to curb the pandemic. One-Dimensional (1D) Convolution Neural Networks (CNN) have been used classify and predict several time-series and sequence data. Here 1D-CNN is applied to the time-series data of confirmed COVID-19 cases for all reporting countries and territories. The model performance was 90.5% accurate. The model was used to develop an automated AI tracker web app (AI Country Monitor) and is hosted on https://aicountrymonitor.org. This article also presents a novel concept of pandemic response curves based on cumulative confirmed cases that can be use to classify the stage of a country or reporting territory. It is our firm believe that this Artificial Intelligence COVID-19 tracker can be extended to other domains such as the monitoring/tracking of Sustainable Development Goals (SDGs) in addition to monitoring and tracking pandemics.

We overcome current limitations in Acute Myeloid Leukemia (AML) diagnosis by leveraging a transfer learning approach from Acute Lymphoblastic Leukemia (ALL) classification models, thus addressing the urgent need for more accurate and accessible AML diagnostic tools. AML has poorer prognosis than ALL, with a 5-year relative survival rate of only 17-19% compared to ALL survival rates of up to 75%, making early and accurate detection of AML paramount. Current diagnostic methods, rely heavily on manual microscopic examination, and are often subjective, time-consuming, and can suffer from inter-observer variability. While machine learning has shown promise in cancer classification, its application to AML detection, particularly leveraging the potential of transfer learning from related cancers like Acute Lymphoblastic Leukemia (ALL), remains underexplored. A comprehensive review of state-of-the-art advancements in acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) classification using deep learning algorithms is undertaken and key approaches are evaluated. The insights gained from this review inform the development of two novel machine learning pipelines designed to benchmark effectiveness of proposed transfer learning approaches. Five pre-trained models are fine-tuned using ALL training data (a novel approach in this context) to optimize their potential for AML classification. The result was the development of a best-in-class (BIC) model that surpasses current state-of-the-art (SOTA) performance in AML classification, advancing the accuracy of machine learning (ML)-driven cancer diagnostics. Author summaryAcute Myeloid Leukemia (AML) is an aggressive cancer with a poor prognosis. Early and accurate diagnosis is critical, but current methods are often subjective and time-consuming. We wanted to create a more accurate diagnostic tool by applying a technique called transfer learning from a similar cancer, Acute Lymphoblastic Leukemia (ALL). Two machine learning pipelines were developed. The first trained five different models on a large AML dataset to establish a baseline. The second pipeline first trained these models on an ALL dataset to "learn" from it before fine-tuning them on the AML data. Our experiments showed that the models that underwent transfer learning process consistently performed better than the models trained on AML data alone. The MobileNetV2 model, in particular, was the best-in-class, outperforming all other models and surpassing the best-reported metrics for AML classification in current literature. Our research demonstrates that transfer learning can enable highly accurate AML diagnostic models. The best-in-class model could potentially be used as a AML diagnostic tool, helping clinicians make faster and more accurate diagnoses, improving patient outcomes.

Diabetes Mellitus (DM) is a metabolic disorder characterized by hyperglycemia, with type 1 characterized as an autoimmune destruction of pancreatic beta cells and type 2 characterized by insulin resistance with progressive beta cell dysfunction. This study applied an existing binary classification algorithm (ALTARN) to accurately predict DM. ALTARN, as a tabular attention residual neural network, uses residual connection to find complex patterns present in tabular columns. We achieved an average training accuracy of 75.22%. Furthermore, a robust set of validation metrics was obtained via five-fold stratified cross-validation, yielding an average accuracy of 74.61%, an average precision of 72.36%, a mean recall of 79.69%, and a mean F1 score of 75.83%.

AO_SCPLOWBSTRACTC_SCPLOWDynamic distribution shifts caused by evolving diseases and demographic changes require domain-incremental adaptation of clinical deep learning models. However, this process of adaptation is often accompanied by catastrophic forgetting, and even the most sophisticated methods are not good enough for clinical applications. This paper studies incremental learning from the perspective of mode connections, that is, the low-loss paths connecting the minimisers of neural architectures (modes or trained weights) in the parameter space. The paper argues for learning the low-loss paths originating from an existing mode and exploring the learned paths to find an acceptable mode for the new domain. The learned paths, and hence the new domain mode, are a function of the existing mode. As a result, unlike traditional incremental learning, the proposed approach is able to exploit information from a deployed model without changing its weights. Pre-COVID and COVID-19 data collected in Oxford University hospitals are used as a case study to demonstrate the need for domain-incremental learning and the advantages of the proposed approach.

Deep learning models have proven the potential to aid professionals with medical image analysis, including many image classification tasks. However, the scarcity of data in medical imaging poses a significant challenge, as the limited availability of diverse and comprehensive datasets hinders the development and evaluation of accurate and robust imaging algorithms and models. Few-shot learning approaches have emerged as a potential solution to address this issue. In this research, we propose to deploy the Generalized Metric Learning Model for Few-Shot X-ray Image Classification. The model comprises a feature extractor to embed images into a lower-dimensional space and a distance-based classifier for label assignment based on the relative distance of these embeddings. We extensively evaluate the model using various pre-trained convolutional neural networks (CNNs) and vision transformers (ViTs) as feature extractors. We also assess the performance of the commonly used distance-based classifiers in several few-shot settings. Finally, we analyze the potential to adapt the feature encoders to the medical domain with both supervised and self-supervised frameworks. Our model achieves 0.689 AUROC in 2-way 5-shot COVID-19 recognition task when combined with REMEDIS (Robust and Efficient Medical Imaging with Self-supervision) domain-adapted model as feature extractor, and 0.802 AUROC in 2-way 5-shot tuberculosis recognition task with domain-adapted DenseNet-121 model. Moreover, the simplicity and flexibility of our approach allows for easy improvement in the feature, either by incorporating other few-shot methods or new, powerful architectures into the pipeline.

Although there is no universal definition for texture, the concept in various forms is nevertheless widely used and a key element of visual perception to analyze images in different fields. The present works main idea relies on the assumption that there exist representative samples, which we refer to as references as well, i.e., "good or bad" samples that represent a given dataset investigated in a particular data analysis problem. These representative samples need to be accounted for when designing predictive models with the aim of improving their performance. In particular, based on a selected subset of texture gray-level co-occurrence matrices (GLCMs) from the training cohort, we propose new representative spatial texture features, which we incorporate into a supervised image classification pipeline. The pipeline relies on the support vector machine (SVM) algorithm along with Bayesian optimization and the Wasserstein metric from optimal mass transport (OMT) theory. The selection of the best, "good and bad," GLCM references is considered for each classification label and performed during the training phase of the SVM classifier using a Bayesian optimizer. We assume that sample fitness is defined based on closeness (in the sense of the Wasserstein metric) and high correlation (Spearmans rank sense) with other samples in the same class. Moreover, the newly defined spatial texture features consist of the Wasserstein distance between the optimally selected references and the remaining samples. We assessed the performance of the proposed classification pipeline in diagnosing the corona virus disease 2019 (COVID-19) from computed tomographic (CT) images.

Leptin is a fat cell-derived hormone involved in satiety and body weight regulation. It also plays a critical regulatory role in the insulin-glucose regulatory system by modulating glucose metabolism and energy homeostasis. However, existing insulin-glucose models often fail to consider the impact of body weight indicators mainly body mass index (BMI) and plasma leptin. To address this limitation, we propose augmenting the ordinary differential equations (ODE) of the Oral Minimal Model (OMM) with an additional equation, incorporating leptin as well as supplementary terms and parameters. By estimating the model parameters, the model behaviour is aligned with the observed data of glucose, insulin and leptin for individuals with type 2 diabetes mellitus (T2DM). Based on model behaviour, revised indices formulated from Oral Glucose Tolerance Test (OGTT) data by including BMI and fasting leptin values are found to have a better correlation with existing indices. Additionally, parameter sensitivity analysis is performed to investigate the influence of the model parameters on the observed variables. Validation of the augmented model with clinical data (without leptin) demonstrates a superior fit to glucose and insulin data compared to the base model. This model emphasizes the intricate associations between leptin, glucose, and insulin concentrations with a potential for developing targeted interventions and therapies for T2DM. Notably, this manuscript introduces the first ODE-based model that incorporates leptin and BMI in the insulin-glucose pathway.