Genomic language models (gLMs) are rapidly becoming important tools for learning biological information directly from sequence data. By adapting concepts from natural language processing, these models aim to capture contextual dependencies, regulatory grammar, evolutionary constraint, and sequence-level functional patterns that may be difficult to detect using alignment-based, motif-based, or conventional supervised methods alone. This systematic review evaluates recent model-development studies of genomic, RNA, nucleotide, codon-level, and regulatory DNA language models, with emphasis on model architecture, tokenization, training objective, biological task, benchmarking strategy, and reported limitations. A structured search of PubMed, Scopus, and Web of Science identified 469 records. After duplicate removal, screening, and full-text eligibility assessment, 58 studies met the strict inclusion criteria for primary model development or substantial model adaptation. The included studies covered diverse applications, including regulatory sequence prediction, variant-effect modeling, genome annotation, microbial and viral genome analysis, RNA splicing and regulation, codon optimization, mRNA design, and generative design of regulatory or RNA sequences. Across the included studies, stronger evidence for gLM utility was generally associated with biologically informed or task-aligned model design, including evolutionary alignments, motif-aware objectives, long-context architectures, RNA structural priors, population-aware representations, and domain-specific pretraining. However, the evidence was heterogeneous and did not support a general claim of superiority over established bioinformatics tools or specialized supervised models. In several regulatory genomics tasks, specialized supervised models, k-mer-based approaches, or conventional deep-learning baselines remained competitive or superior to pretrained language-model representations. Generative models showed growing promise for RNA, codon, viral genome, and cis-regulatory element design, although many were evaluated mainly in silico. Overall, the field is advancing quickly, but broader impact will require standardized benchmarks, clearer reporting, stronger external validation, improved interpretability, and experimental confirmation of predicted or generated biological functions.

The surgical suite (SS) is one of the most critical and resource-intensive hospital departments. Inefficient spatial arrangements can increase travel distances, reduce workflow efficiency, and negatively affect medical staff performance and patient care. However, designing an optimal SS layout is a challenging task due to the complex relationships among rooms, personnel, equipment, and clinical activities. This study addresses the problem of SS layout optimization by developing a multi-objective mixed-integer linear programming (MILP) model. The aim is to simultaneously minimize the intra-suite travel costs and maximize functional adjacency among rooms. The optimization framework was implemented using ILOG CPLEX 12.8 and MATLAB R2022a. To validate the proposed approach, two room configurations comprising thirteen and nine functional spaces were investigated through a case study of a newly planned Egyptian private hospital. This work integrates an adjacency index within the developed framework to improve compliance with healthcare design requirements. Different optimization scenarios and adjacency constraints were evaluated to assess their influence on layout quality and computational performance. The findings demonstrated that incorporating adjacency constraints significantly improved room arrangements and produced conceptually optimized layouts. The nine-room configuration achieved superior computational performance and more practical layouts than the thirteen-room configuration. The proposed framework provides healthcare planners with a decision-support tool for generating an efficient SS layout, reducing reliance on trial-and-error design approaches.

Challenges in healthcare facilities, especially emergency departments (EDs), persistently arise from increasing patient volumes, pandemics, and complex operational factors. This study explores the use of graph-theoretic methods as powerful tools for optimizing hospital layouts. Initially, we introduced a multi-objective optimization framework informed by graphs for the design of ED layouts, utilizing NSGA-II and GDE3, with the objectives of minimizing patient flow costs and enhancing the spatial proximity of service areas. We propose and confirm the application of both local and global graph-theoretic metrics, including centrality metrics, clustering coefficients, and network efficiency, to assess and rank Pareto-optimal layouts. Simulation results on the layout of an ED in Dalian, China, show that the best-ranked layouts exhibit superior graph-theoretic values, with the NSGA-II and GDE3 solutions reducing patient flow cost by 18.32% and 11.42%, respectively, and improving service area closeness by 14.5% and 18.02%, respectively. Subsequently, we present a layout prediction method based on machine learning that employs multi-output regression models trained on graph-theoretic features extracted from a large set of optimization-based layout solutions. The proposed models, particularly decision trees and random forests, demonstrated impressive prediction accuracies of up to 98.18% and 88.06%, respectively, showcasing their effectiveness in efficiently producing optimal layouts. Collectively, these findings highlight the potential of graph-theoretic models in enhancing hospital design and operational decision-making.

This study presents a hybrid Multi-Criteria Decision Making (MCDM) and Machine Learning (ML) framework designed to support operational decisions related to the upgrade or replacement of medical imaging systems. Focusing on Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) units installed across healthcare facilities in Egypt, the research offers a novel contribution by analyzing equipment lifecycle management from the manufacturer's perspective, rather than that of the healthcare provider. A real-world dataset from Siemens Healthineers was used, incorporating six performance indicators such as system age, utilization rate, and downtime. Two sets of weights were applied-objective CRiteria Importance Through Intercriteria Correlation (CRITIC) and expert-based (Service Marketing Head)-within the Evaluation based on Distance from Average Solution (EDAS) method to generate replacement prioritization rankings. Furthermore, four supervised ML algorithms were trained to classify systems into three lifecycle stages. This hybrid approach demonstrated a classification accuracy of up to 96.5 % in identifying these classes. Results highlight the value of integrating domain expertise and data-driven techniques for supporting lifecycle planning, installed base retention, and service strategy development in the medical imaging industry.

Imaging genetics is one of the important keys to precision medicine that leads to personalized treatment based on a patient's genetics, phenotype, or psychosocial characteristics. It deepens the understanding of the mechanisms through which genetic variations contribute to neurological and psychiatric disorders. This systematic review overviews the methods and applications of imaging genetics in the context of neurological diseases, mentioning its potential role in personalized medicine. Following PRISMA guidelines, this review systematically analyzes 28 studies integrating genetic and neuroimaging data to explore disease mechanisms and their implications for precision medicine. Selected research included multiple neurological disorders, including frontotemporal dementia, Alzheimer's disease, bipolar disorder, schizophrenia, Parkinson's disease, and others. Voxel-based morphometry was the most common imaging technique, while frequently examined genetic variants included APOE, C9orf72, MAPT, GRN, COMT, and BDNF. Associations between these variants and regional gray matter loss (e.g., frontal, temporal, or subcortical regions) suggest that genetic risk factors play a key role in disease pathophysiology. Integrating genetic and neuroimaging analyses enhances our understanding of disease mechanisms and supports advancements in precision medicine.

RNA gene expression datasets are inherently high-dimensional and often suffer from class imbalance and redundancy. In this study, we propose a robust framework that combines embedded feature selection, three techniques, with a deep Transformer Encoder Convolutional Decoder (TECD)-based embedding model. Synthetic Minority Oversampling Technique (SMOTE) is applied to address class imbalance during training. RNA-seq dataset of matched tumor tissues (n=360) grouped into four Triple-Negative Breast Cancer (TNBC) demonstrates that TECD reduces dimensionality by over 99% while enhancing performance across eight various classifiers. TECD enabled simple models like K-NN and Logistic Regression to outperform more complex models like XGBoost, with K-NN improving from 63.9% to 93.1% accuracy, and SVM with TECD achieving the highest accuracy of 94.44%. This highlights TECD’s strength in producing highly informative representations even without relying on complex classifier architectures.

One of the most intricate parts of the human body is the spine. It serves multiple purposes. It supports and helps the movement of the body. It carries the weight of the entire abdomen, the upper limbs, and the head. The aim of this study is to establish a feasible 3D finite element model of the lumbar spine using the finite element method route and validate it by comparing the simulation results with data from in vitro experiments. A lumbar spine (L1–L5) finite element (FE) model was developed, and it included posterior fixation of pedicle screws (PS) at the L3–L4 segment level. This FE study investigated the impact of the posterior PS fixation system on the lumbar spine’s biomechanics using different materials for the screw-rod fixation system. Using titanium and CFR-PEEK materials, the impact of a posterior PS fixation system on lumbar spine biomechanics was examined for all physiological motions. The CFR-PEEK fixation system showed a reduction in von Misses stress at all physiological motions and an increase in the range of motion, which will increase the patient’s daily life performance rate and decrease the possibility of screw loosening and adjacent segment degeneration. The study concludes that CFR-PEEK rods are an alternate rod material to prevent the drawbacks of rigid-type rod fixation. CFR-PEEK implants have excellent mechanical stability and load-bearing capacity, reducing the likelihood of implant failure and promoting effective fusion. Results demonstrate how CFR-PEEK rods may lessen implant-related issues such as adjacent segment degeneration and screw loosening. Clinically, this could result in better long-term results for patients having lumbar fusion, decreased rates of revision surgery, and increased postoperative mobility.

Triple-negative breast cancer (TNBC) characterizes a significant clinical challenge due to limited therapeutic options resulting from the nonexistence of hormone receptors and HER2. The Basal-Like Immune-Suppressed (BLIS) subtype exhibits intensely poor outcomes due to immune avoidance mechanisms. This study employs inclusive bioinformatics approaches to recognize immune-related hub genes within the BLIS subtype to uncover potential therapeutic targets and prognostic biomarkers. Starting with the gene expression dataset containing 58,000 genes from 360 TNBC patients, we filtered low-expression genes and applied variance stabilizing transformation (VST) using DESeq2. Differential expression analysis across the four recognized TNBC subtypes—BLIS, Mesenchymal (MES), Luminal Androgen Receptor (LAR), and Immunomodulatory (IM) identified 353 significantly expressed genes, comprising 124 upregulated and 229 downregulated differentially expressed genes (DEGs). Pathway enrichment analysis revealed significant dysregulation of immune-related processes. We constructed a protein–protein interaction (PPI) network with 36 genes and applied Density-Based Spatial Clustering of Applications with Noise (DBSCAN) in STRING with a high confidence threshold (0.900). Using Cytoscape based on the Matthews Correlation Coefficient (MCC) method, we identified ten hub genes with the highest network connectivity: CXCR3, CXCL10, IFNG, CCL5, CXCL9, CCR5, CX3CL1, CCL11, CCL4, and CXCL11. Focusing on downregulated immune-related hub genes in the BLIS subtype, Kaplan–Meier survival analysis based on relapse-free survival (RFS) and subsequent multivariate analysis identified CCR5 and IFNG as novel biomarkers significantly associated with survival outcomes. Our findings provide a foundation for developing immune-targeted therapeutic approaches for BLIS-TNBC patients and provision the integration of machine learning models to predict treatment responses and optimize patient-specific strategies, potentially transforming the clinical management of this challenging breast cancer subtype.

The prevalence of vision impairment is increasing at an alarming rate. The goal of the study was to create an automated method that uses optical coherence tomography (OCT) to classify retinal disorders into four categories: choroidal neovascularization, diabetic macular edema, drusen, and normal cases. This study proposed a new framework that combines machine learning and deep learning-based techniques. The utilized classifiers were support vector machine (SVM), K-nearest neighbor (K-NN), decision tree (DT), and ensemble model (EM). A feature extractor, the InceptionV3 convolutional neural network, was also employed. The performance of the models was evaluated against nine criteria using a dataset of 18000 OCT images. For the SVM, K-NN, DT, and EM classifiers, the analysis exhibited state-of-the-art performance, with classification accuracies of 99.43%, 99.54%, 97.98%, and 99.31%, respectively. A promising methodology has been introduced for the automatic identification and classification of retinal disorders, leading to reduced human error and saved time.

ABSTRACTPapilledema is a prevalent neuro‐ophthalmic condition characterized by optic disk swelling. It is known to pose a significant risk of vision loss in its advanced stages. To address the pressing need for accurate detection and grading of papilledema, this study introduces a novel approach utilizing optical coherence tomography (OCT) scans. A cascaded model that combines four transfer learning models—SqueezeNet, AlexNet, GoogleNet, and ResNet‐50—for both the detection and grading phases was proposed. Additionally, a specialized convolutional neural network (CNN) model is meticulously designed to cater specifically to the complexities of papilledema analysis. Unlike the fundus camera‐based models, this study integrates deep learning models for the diagnosis of papilledema from OCT scans. A new dataset of OCT scans was collected to ensure a comprehensive evaluation of the models. It encompasses a wide range of papilledema, pseudopapilledema, and normal cases. This dataset serves as a valuable resource for training and testing of the proposed models. In addition, two validation strategies have been adopted to ensure the model's generalizability and robustness. Furthermore, it enhances the model's accuracy and reliability. The results are highly promising; remarkable accuracy rates have been achieved. Specifically, the SqueezeNet, AlexNet, GoogleNet, ResNet‐50, and customized CNN models achieved accuracy levels of 98.44%, 98.50%, 98.28%, 98.30%, and 96.26%, respectively, for the handout validation strategy. These findings not only demonstrate the efficacy of using deep learning in papilledema detection and grading but also establish the superiority of the proposed models when compared with other relevant studies. By addressing the challenges associated with papilledema, the study significantly contributes to the advancement of neuro‐ophthalmic diagnostics. The accurate and efficient detection of papilledema from OCT scans holds immense potential for guiding timely interventions and preserving patients' visual health.

The complexities inherent in diagnosing papilledema, particularly within the realm of neuro-ophthalmology, emphasize the pressing need for sophisticated diagnostic methodologies. This study highlights the application of novel models tailored explicitly for papilledema detection, distinguishing it from pseudo-papilledema and normal cases, through the strategic utilization of deep learning frameworks, specifically convolutional neural networks (CNNs) and recurrent neural networks (RNNs). Leveraging hierarchical feature extraction from retinal images, the multi-paths CNN model successfully identifies crucial indicators of papilledema, while the cascaded model, integrating ResNet-50 and long short-term memory (LSTM), effectively captures sequential features. Leveraging a meticulously curated dataset comprising 18,258 fundus images, the presented models exhibit exceptional performance, with the multi-paths CNN achieving an accuracy of 99.97 %, and the LSTM model demonstrating an accuracy of 99.81 %. Comparative analysis showcases the unparalleled efficacy of the models, underscoring their potential in clinical diagnostics. Notably, they demonstrate robustness in occlusion sensitivity tests, highlighting their resilience in scenarios involving obscured image components. This pioneering study represents a significant milestone in papilledema detection, with the promise of advancing patient outcomes and streamlining healthcare practices. The proposed deep learning models not only offer precise diagnoses but also hold the potential to automate elements of the diagnostic workflow, alleviating the workload of healthcare professionals and enhancing overall patient care outcomes.

One of the most widely used imaging modalities for evaluating the structure and function of the heart is echocardiography. Many features related to the heart's four chambers are driven by echocardiography to diagnose heart health. The left ventricle ejection fraction (LVEF) is one of the most influential characteristics considering the left ventricle. This study uses echocardiographic videos from the EchoNet-Dynamic dataset to evaluate left ventricular function, focusing on apical four-chamber (A4C) views. Sequential frame extraction from these videos enabled the application of novel methods yielding precise results. Extracted features were employed in the Gaussian Process Regression (GPR) model to predict LVEF, demonstrating superior performance (Root Mean Square Error (RMSE)= 3.66, R 2 = 0.957) compared to other methods explored. These findings underscore the potential of our proposed methodology in enhancing cardiac function assessment through echocardiographic analysis.

AbstractPlanning and budgeting for healthcare facilities necessitate precise estimation and a thoughtful analysis of the required area for each department, as well as extensive knowledge of the most important elements that can influence the design and provide an appropriate environment for staff and patients. A gross conversion factor is influential in calculating the entire area of a department. However, there is no precise standard indicating how this factor should be calculated. To address this problem, quality function deployment (QFD) has been implemented. The intensive care unit (ICU) is one of the highly equipped departments that should be properly planned; therefore, it was selected for application. The five steps of the QFD are used to identify and rank the most important and efficient design features. Only the top factors are considered in calculating the required area for the ICU department. This approach presents a unique method for estimating the gross square area based on calculating the gross conversion factor. To demonstrate the proposed approach’s efficacy, it was applied to two Egyptian public hospitals. The results reveal that 15–16% of the total space of both hospitals can be reduced from the total department space. Moreover, an outstanding reduction in ICU design and preparation expenses has been achieved. The design reduces the required budget for a project while maintaining the same construction material quality and accommodating extra beds. Additionally, the QFD proves its ability to redesign the ICU department while reducing costs.

Papilledema is characterized by optic disc swelling due to increased intracranial pressure, necessitates accurate diagnostic tools for effective treatment planning. This study presents an innovative artificial intelligence (AI) model employing fuzzy logic to grade papilledema severity. The study utilized a newly collected dataset containing different clinical measurements. The input of the system includes Best-Corrected Visual Acuity (BCVA), pupil reaction, retinal nerve fiber layer (RFNL) thickness, total peripapillary thickness, and visual field measurements and output of the system is to identify the level of severity of papilledema within four grades. The fuzzy logic model achieves a classification accuracy exceeding 90%, demonstrating its effectiveness in discerning diverse papilledema severity levels. Comparative analysis with existing literature highlights competitive performance, positioning the model as a promising neuro-ophthalmological diagnostic tool. Unique contributions include dataset composition, feature extraction, and the model's implementation. This study pioneers in advancing diagnostic methodologies for papilledema, offering a reliable tool for ophthalmologists used for the first time in papilledema severity detection. The results indicate the model's potential for integration into clinical practice, contributing AI-driven neuro-ophthalmological diagnostics and enhancing patient care.

Lung cancer has a high incidence rate and is considered highly fatal because of its low survival rate at early stages compared to other cancers. Computed tomography (CT) scans can reveal pulmonary nodules of different shapes and volumes in two dimensional (2D) slices. Three-dimensional (3D) reconstruction of pulmonary nodules can assist the radiologist in early treatment appropriate for the 3D nodule volume screened. In this research, we present a 3D reconstruction algorithm that uses 2D CT slices to reconstruct a 3D lung nodule. The equivalent diameters of small nodules ranged from 3 to 30 mm. A segmentation approach (based on bounding boxes and maximum intensity projection) was applied. Extracting the lung nodules from the 2D candidate masses was performed via a rule-based classifier. Surface rendering was used to reconstruct 3D pulmonary nodules which were visualized on the 3D Slicer software. The 3D nodule volume, as well as the accuracy rate and error of volume estimation were calculated. The proposed methodology was validated against the actual volumes of 14 3D nodules from the Lung Image Database Consortium (LIDC) database. The proposed algorithm achieved a maximum accuracy of 99.6627 % for lung nodule volume estimation. The corresponding average accuracy rate and average percentage error were 97.34 % and 2.66 %, respectively. The screening of 3D lung nodules can support surgery planning via nodule volume estimation. The average accuracy and error rates of the 3D reconstruction algorithm showed promising results in comparison with other published studies.

With the increasing availability of digital X-ray imaging, artificial intelligence (AI) has emerged as a promising tool for automating the assessment of thoracic diseases. The objective of this study is to systematically review the artificial intelligence (AI) and deep learning methods proposed for the automated assessment of thoracic diseases from chest X-ray images. A thorough search of the relevant literature was conducted, and studies that met the inclusion criteria were critically reviewed. Information on the datasets, model architectures, evaluation metrics, and results was extracted. Convolutional neural networks are prevalent, achieving a state-of-the-art classification performance. Recent studies have explored more complex tasks such as disease localization, segmentation, and report generation. The multitask and multimodal approaches are promising. Challenges related to the data, evaluations, and clinical adoption were identified. This study prevails that there is a significant progress in using deep learning for automated chest X-ray analysis. Further research is needed to validate these models in real-world settings and to facilitate their integration into clinical workflows.

Alzheimer's Disease (AD) presents a growing global health challenge, with early diagnosis and intervention being of paramount importance. This research leverages the power of machine learning to classify different stages of AD using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Seven diverse machine learning algorithms, including K-nearest Neighbors (K-NN), Decision Tree (DT), Ensemble Models, Naive Bayes, Neural Network, Support Vector Machine (SVM), and Discriminant Analysis are applied. The output was classifying patients into three distinct AD stages: Alzheimer's Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN). Data preprocessing involves normalizing date formats, addressing missing data, and encoding diagnostic categories. The results highlight the effectiveness of these algorithms in accurately classifying AD stages, with the ensemble model achieving a remarkable 97.70% accuracy on the test dataset. Decision Tree, Neural Network, K-NN, and SVM also demonstrate strong performance, showcasing their potential in clinical applications.

Outlier detection (OD) is a key problem, for which numerous solutions have been proposed. To deal with the difficulties associated with outlier detection across various domains and data characteristics, ensembles of outlier detectors have recently been employed to improve the performance of individual outlier detectors. In this paper, we follow an ensemble outlier detection approach in which good outlier detectors are selected through an enhanced clustering-based dynamic selection (CBDS) method. In this method, a bisecting K-means clustering algorithm is employed to partition the input data into clusters where every cluster defines a local region of competence. Among the initial pool of detectors, the outputs of the detectors with the most competent local performance were combined through four possible schemes to produce the final OD results. Experimental evaluation and comparison of our method were carried out against four variants of locally selective combination in parallel (LSCP) outlier ensembles. The CBDS-based schemes compare well with the LSCP-based ones on 16 public benchmark datasets and incur considerably lower computational costs. The CBDS method consistently achieved superior average scores of the area under the curve (AUC) of the receiver operating characteristic (ROC), and particularly outperformed the LSCP method on nine of the 16 datasets in terms of the AUC score. In addition, while the CBDS and LSCP methods have similar computational costs on small datasets, the CBDS method achieves significant time savings compared with the LSCP method on large datasets.

Facial paralysis (FP) is an inability to move facial muscles voluntarily, affecting daily activities. There is a need for quantitative assessment and severity level classification of FP to evaluate the condition. None of the available tools are widely accepted. A comprehensive FP evaluation system has been developed by the authors. The system extracts real-time facial animation units (FAUs) using the Kinect V2 sensor and includes both FP assessment and classification. This paper describes the development and testing of the FP classification phase. A dataset of 375 records from 13 unilateral FP patients and 1650 records from 50 control subjects was compiled. Artificial Intelligence and Machine Learning methods are used to classify seven FP categories: the normal case and three severity levels: mild, moderate, and severe for the left and right sides. For better prediction results (Accuracy = 96.8%, Sensitivity = 88.9% and Specificity = 99%), an ensemble learning classifier was developed rather than one weak classifier. The ensemble approach based on SVMs was proposed for the high-dimensional data to gather the advantages of stacking and bagging. To address the problem of an imbalanced dataset, a hybrid strategy combining three separate techniques was used. Model robustness and stability was evaluated using fivefold cross-validation. The results showed that the classifier is robust, stable and performs well for different train and test samples. The study demonstrates that FAUs acquired by the Kinect sensor can be used in classifying FP. The developed FP assessment and classification system provides a detailed quantitative report and has significant advantages over existing grading scales.

DNA microarray data sets have been widely explored and used to analyze data without any previous biological background. However, analyzing them becomes challenging if data are missing. Thus, machine learning techniques are applied because microarray technology is promising in genomics, especially in the analysis of gene expression data. Furthermore, gene expression data can describe the transcription and translation processes of each genetic information in detail. In this study, a new system was proposed to impute more realizable values for missing data in a microarray dataset. This system was validated and evaluated on 42 samples of rectal cancer. Several evaluation tests were also conducted to confirm the effectiveness of the new system and compare it with highly known imputing algorithms. The proposed clustering column-mean quantile median technique could predict highly informative missing genes, thereby reducing the difference between the original and imputed datasets and demonstrating its efficiency.