Artifical Intelligence

Coronary artery disease (CAD) affects an estimated 11.39 million people in China, and mortality has continued to rise since 2012. Quality-of-care data indicate persistent gaps in diagnostic and therapeutic guideline adherence, incomplete implementation of secondary prevention, and suboptimal risk-factor control. Large language models (LLMs) can rapidly retrieve and integrate literature, medical records, and multimodal clinical data, and may support clinical decision-making. However, most existing medical LLMs are general-purpose, are trained predominantly on published literature and web text rather than authentic clinical records, and have not been validated with real patient data. No disease-specific LLM for CAD currently exists, and no evaluation framework tailored to CAD has been established; existing benchmarks are general-purpose, lack specialty-specific risk grading, are not rigorously anchored to current guidelines, and emphasize literal accuracy over clinically critical reasoning steps and prescription concordance. This prospective, multicenter, observational study will enroll approximately 640 patients with confirmed or suspected CAD at Fuwai Hospital and collaborating centers. No study-specific intervention, examination, biospecimen collection, or additional follow-up visit is performed; all clinical decisions are made independently by the treating physician according to routine standards of care. For consenting participants, routinely generated clinical documentation from the index outpatient visit or hospitalization - including medical history, laboratory results, imaging and coronary angiography/coronary CT angiography reports, physician notes, online consultation records, and follow-up records - will be de-identified and used to construct a high-quality multimodal CAD dataset. This dataset will support the development of a CAD-specific large language model (CorAI), the construction of an evaluation framework spanning eight predefined clinical scenarios and multiple assessment dimensions, and external validation of the model at participating centers. The primary objectives are to characterize the completeness and quality of the constructed prospective cohort dataset, and to quantify the guideline concordance of CorAI-generated clinical recommendations as adjudicated by a blinded panel of cardiologists.

Some people have a "gummy smile," where too much gum tissue shows when they smile. Doctors currently diagnose this by hand, using rulers and visual judgment. This study is testing whether a computer program using artificial intelligence (AI) can help identify and measure gummy smiles from photographs, as accurately as expert dentists do it manually. Patients seen at Cairo University's Periodontics and Orthodontics clinics will have photos of their smile taken. Researchers will use these photos to first "teach" the AI program how to recognize and measure a gummy smile. Then, the AI will be tested on a new set of patient photos, and its measurements will be compared to measurements made by expert dentists, to see how closely they match. The goal is to find out whether this AI tool could one day help doctors diagnose gummy smiles more quickly and consistently, which may improve care for patients.

BEACON (Benchmarking AI for Clinical Oncology decisioNmaking) is a prospective, multicentre, comparative, blinded, non-interventional benchmark evaluating the treatment recommendations of five frontier large language models (LLMs) against the recommendations of multidisciplinary tumour boards (RCP) in oncology treatment planning. One hundred standardised synthetic cases (20 per localisation, across breast, lung, urological, digestive and gynaecological cancers) are submitted as identical structured input to two independent tumour boards per localisation and to five frontier LLMs. Each recommendation - human or model - is decomposed into five predefined decision domains (intent, surgery, radiotherapy, systemic therapy, work-up and biomarkers) and scored 0/1/2 for concordance against a two-tier reference: the consensus of the two tumour boards, complemented by an a priori locked guideline matrix (ESMO, NCCN). The primary endpoint is domain-level concordance between LLM and RCP consensus, expressed as a linearly weighted Cohen's kappa. A co-primary safety endpoint captures the proportion of recommendations carrying serious harm potential, because concordance alone can conceal dangerous errors. Because expert boards may disagree with one another on identical cases, model performance is always interpreted against the human consensus. BEACON is designed as reusable, openly licensed, pre-registered infrastructure: all synthetic cases, evaluation rubrics, the locked guideline matrix, scoring algorithms and verbatim prompts are released for full reproducibility.

This retrospective validity study evaluates the accuracy of artificial intelligence (AI) in determining gender from mandibular morphometric linear measurements. The study utilizes pre-existing Cone Beam Computed Tomography (CBCT) scans of adult Egyptian dental patients. After automatic segmentation of the mandible from these scans, a radiologist will manually perform measurements from certain anatomical points. These measurements will be the reference standard for the AI models. A three-dimensional deep learning model will be developed to perform two tasks: 1. To identify the anatomical points and make the specified linear measurements from these points on the segmented mandibles 2. To accurately predict gender based on these measurements. (Main Objective) The primary objective of this study is to evaluate the accuracy of machine learning algorithms in gender identification from linear morphometric measurements of the mandible. The known gender from patient records will serve as the reference standard. This study will assess the reliability of AI as an objective tool for gender determination for forensic purposes.

This study evaluates whether large language models (LLMs) can reliably assess written clinical-reasoning case examinations completed by undergraduate physiotherapy students, compared with faculty assessment. In the course "Specific Methods in Physiotherapy" (third year of the Physiotherapy Degree), students solve complex clinical cases that require clinical reasoning, technical knowledge, and therapeutic decision-making. These cases are traditionally graded by faculty, a time-consuming process that may show inter-rater variability. A set of de-identified student case examinations will be assessed using the rubric currently applied in the course, which covers clarity and structure of clinical reasoning, integration of the biopsychosocial model (ICF and APTA frameworks), accuracy in identifying pain mechanisms, coherence between diagnosis, hypotheses, and treatment, originality and depth of analysis, and professional writing. Each examination will be scored independently by three LLMs (for example, Claude, ChatGPT, and Gemini), each receiving an identical standardized prompt that embeds the same rubric, and by faculty serving as the reference standard. To avoid overloading faculty, full double human grading may not be feasible; the human reference will therefore consist of expert faculty grading by one independent rater or, when resources allow, two independent raters. In contrast, paired assessment is fully implemented across the AI models: each examination is scored by several LLMs, and each model is queried in duplicate, allowing the study to estimate agreement between models and the test-retest stability of each model. The primary aim is to quantify agreement between LLM-generated scores and the faculty reference score. Secondary aims include agreement among the LLMs, test-retest reliability of each model, criterion-level agreement, the quality and usefulness of the qualitative feedback generated, the time and cost associated with each approach, and students' perceptions of the usefulness of human versus AI feedback. The findings will clarify the strengths and limitations of LLMs as supportive tools for formative assessment in health-professions education and will inform criteria for their responsible and effective use. No LLM output will affect students' official grades, which remain the sole responsibility of faculty.

This study aims to interpret the Bispectral Index (BIS) monitoring method, which we routinely use for monitoring in scoliosis surgery, with artificial intelligence (AI) tools and to determine the accuracy and reliability of AI tools in clinical practice by comparing this interpretation with the interpretations of two clinicians experienced in BIS.

This is a multi-center, cross-sectional study evaluating a smartphone-based artificial intelligence (AI) system for anterior segment eye disease screening. The system is designed to identify 16 clinically important anterior segment conditions from images captured using a standard Android smartphone. A core design feature of the system is that all image analysis is performed entirely on the smartphone itself, without requiring internet connectivity or cloud-based server infrastructure. The study is motivated by a structural challenge in the deployment of medical AI: systems that depend on cloud infrastructure for inference are non-functional in settings without reliable internet access, which disproportionately excludes populations in low-resource regions where the burden of preventable eye disease is highest. This study evaluates whether an on-device AI system, designed with operational constraints as a primary engineering objective, can deliver clinically acceptable diagnostic performance while remaining operable under real-world connectivity limitations. The study comprises five evaluation components. First, the diagnostic performance of the AI system is benchmarked against board-certified ophthalmologists of varying seniority on a standardized set of smartphone-captured anterior segment images. Second, the usability of the system is evaluated among non-medical users who perform self-administered screening with minimal instruction, with per-screening time recorded across consecutive attempts to characterize the learning curve. Third, a head-to-head field trial directly compares the on-device AI system against a functionally equivalent cloud-based deployment of the same model architecture across key operational dimensions including screening duration, diagnostic performance, and user acceptability. Fourth, population-level screening is conducted among consecutively enrolled community residents at two low-resource sites, with per-disease sensitivity and specificity calculated against reference-standard slit-lamp examinations. Fifth, pre-specified health-economic and environmental analyses compare the two deployment modalities in terms of per-person screening cost, cost-effectiveness, per-inference electricity consumption, and projected carbon emissions at scale. The reference standard for all diagnostic comparisons is slit-lamp biomicroscopic examination performed by board-certified ophthalmologists. The study is designed and reported in accordance with the DECIDE-AI reporting guideline for early-stage clinical evaluation of AI-driven decision-support systems.

Heart valve diseases are among the most serious cardiovascular conditions in older age. One of the most common forms is aortic valve stenosis, a narrowing of the valve opening between the left ventricle and the main artery. As the valve becomes tighter, the heart must work harder and harder to pump blood through the body. This process often develops slowly over many years and initially causes no clear symptoms. As a result, the condition is frequently detected only in advanced stages, when warning signs such as shortness of breath, chest pain, or dizziness appear. Without treatment, aortic valve stenosis can become life-threatening. If detected early, however, very effective treatment options are available today. Up to now, the disease has been reliably diagnosed mainly through echocardiography. Yet this method is complex, costly, and requires specialized medical staff. A simple, affordable, and broadly accessible screening option does not yet exist. The interdisciplinary clinical research project explores whether conventional smartphones could fill this gap. Almost all modern devices are equipped with sensors such as microphones, accelerometers, and gyroscopes. These can capture both heart sounds and subtle vibrations of the chest. The research team is investigating whether reliable diagnostic information for the diagnosis of aortic valve stenosis can be extracted from such recordings. To achieve this, the signals are processed with newly developed methods and analyzed using artificial intelligence. For the study, several hundred patients with and without valve disease will be examined. The smartphone results will be compared with established diagnostic standards, particularly echocardiography, to test accuracy and reliability. If successful, the approach could enable a straightforward, digital heart check at home using nothing more than a conventional smartphone. Such a tool would provide an accessible, low-cost, and widely available method for early detection, helping more people receive timely and potentially life-saving treatment.

Freezing of gait (FOG) is a debilitating symptom of Parkinson's disease increases the risk of falling. Despite being a common symptom, it is still difficult to evaluate freezing of gait quickly and accurately. Currently, the gold-standard method to determine the severity of FOG is a manual analysis of video footage by an experienced assessor, collected during standardized FOG-provoking walking tests. Because this is a very time-intensive process, where different assessors sometimes obtain different results, our team at KU Leuven have developed an artificial-intelligent (AI) algorithm trained to identify FOG episodes based on wearable inertial measurement unit (IMU) sensor data. The AI algorithm has already undergone initial validation during laboratory testing, yielding promising results. The aim of this study is to investigate whether the AI algorithm can accurately detect FOG episodes in a less controlled environment, namely the home environment. In a second phase, the investigators will also use the collected data to improve the AI algorithm for automated FOG detection in the home. Finally, the investigators want to explore whether the AI algorithm can detect FOG in real-time.

The global increase in cancer cases has made breast cancer the second most common cancer after lung cancer and a primary health problem among women. Early diagnosis is the most critical factor in improving survival rates and quality of life in breast cancer. Breast self-examination (BSE), which enables individuals to notice changes in their own breast tissue during the early diagnosis process, is a low-cost and effective awareness method. It is essential that nurses, who play a key role in raising public awareness on this issue, and nursing students, who are the future healthcare professionals, have sufficient knowledge and practical skills in BSE. However, the literature shows that even if students have theoretical knowledge, their application rates are low. In this context, the "AI-Supported Flipped Learning" model, which goes beyond traditional methods and supports active learning, personalized feedback, and digital literacy, has the potential to be an innovative solution in nursing education. Objective: This study aims to evaluate the effect of AI-supported flipped learning model and traditional education on the knowledge levels and performance skills of nursing students regarding BSE knowledge and skills.

The goal of this clinical trial is to develop and evaluate an Artificial Intelligence Assisted Prehabilitation Program (AI APP) for returning to work and cost-effectiveness analysis in patients with oral cancer (OC). The main questions it aims to answer are: * What kinds of needs are related to returning to work (RTW) in patients with OC from diagnosis to survival that we can incorporate into the development of AI APP to assist this population ? * How is the effect of the AI APP that based on findings from the first question for patients with OC on physical and psychological distress, fear of recurrence, self-efficacy in coping with cancer, communication, motor function, quality of life, and RTW? * How is the effect of the RTW AI prediction model to identify high-risk groups ? And how is the comprehensive cost effectiveness of benefits and quality of life of the AI APP for OC population? Researchers will compare patients without using AI APP to see if the AI APP works to assist with coping physical and psychological distress, communication, motor function, quality of life, and RTW issues for individuals with OC? Participants will: * Be asked to fulfill a structural questionnaire, or engage in a semi-structured one-by-one interview or a focus group to assess their physical, psychological, and social support needs in the first stage. * Be invited to participant the pilot testing of AI APP in the second stage. * Be provided and trained by 3-month AI APP for 3 months or cared as usual in the third stage. * Complete a structural questionnaire and follow up one year, including the baseline (before using the AI app) and at 1-2 weeks, 3 months, 6 months, 9 months, and 12 months after the baseline. * Engage in one-by-one interview or a focus group to assess user experiences of the AI APP.

Cardiovascular disease is the leading cause of death worldwide, and individuals with diabetes or other cardiometabolic conditions are at increased risk of adverse cardiovascular outcomes. Although advances in prevention and treatment have reduced cardiovascular events globally, cardiometabolic disease continues to represent a significant health burden, particularly in regions with high diabetes prevalence. In Qatar and other Gulf Cooperation Council countries, the prevalence of diabetes and obesity is increasing, contributing to a high proportion of participants presenting with acute coronary syndrome who have type 2 diabetes or prediabetes. This observational study will use electronic medical record data from patients hospitalized at the Heart Hospital with acute coronary syndrome and a concomitant diagnosis of diabetes or prediabetes. The study will assess trends in cardiovascular risk factors and cardiovascular events, including readmission and mortality. An artificial intelligence component will be used to develop and validate machine learning based risk prediction models to forecast adverse cardiovascular outcomes in participants with cardiometabolic disease. These models will integrate clinical, biochemical, imaging, and other non-invasive data routinely collected during participants care to identify predictors of cardiovascular events.

The integration of Artificial Intelligence (AI) in anesthesiology offers the potential to shift patient monitoring from reactive to predictive. Deep learning architectures, specifically Long Short-Term Memory (LSTM) networks, excel at processing complex, time-series data to forecast future clinical states. While standard PK/PD models (such as the state of the art Eleveld model for Propofol and Remifentanil) estimate target-site drug concentrations (Ce), they do not account for real-time, patient-specific dynamic responses. This study aims to deploy an AI framework designed to predict future physiological states.

The goal of this prospective study is to to evaluate the prevalence of metachronous advanced colonic lesions in subsequent surveillance colonoscopies in patients who had previously undergone AI-assisted colonoscopy to conventional colonoscopy examinations. The main question it aims to answer is whether employing AI-assisted colonoscopy can decrease the likelihood of metachronous advanced colonic lesions during subsequent surveillance colonoscopies. Researchers will compare patient who undergo conventional colonoscopy in previous colonoscopy to see if AI-assisted colonoscopy can decrease the likelihood of metachronous advanced colonic lesions during subsequent surveillance colonoscopies. Participants will undergo surveillance colonoscopy to assess the presence of metachronous advanced colonic lesion

This clinical trial is testing whether plain language summaries made by artificial intelligence help people understand their eye doctor's notes better. Adults receiving eye care at the Jules Stein Eye Institute will get either the usual medical notes or a note with the addition of an AI-generated summary that explains the information in simple, everyday words. Participants will then answer a short survey and receive a follow-up call to share how clear the information was, how well they understood their diagnosis and treatment, and whether they feel more confident about their care. The goal is to find out if these plain language summaries can make it easier for people to understand their eye care and improve communication between patients and health care providers.

This study would answer the following question:Does AI application-based training improve cognitive function in Patients with Stroke? The aims of this study: To investigate the efficacy of AI application-based training on cognitive function in stroke patients.

This study seeks to validate the real-world accuracy of an AI-based algorithm for identifying the location of an accessory pathway from the 12-lead electrocardiogram

Retrospectively collect the clinical data, breast MRI images, breast ultrasound images and reports, laboratory indicators (such as CA199, CA153, CA125, CEA/AFP), pathological diagnosis results, HE staining images, and existing immunohistochemical results (including CD8A, KPT5, GFRA1, PFKP, ER/PR percentage, Her-2 expression, Ki-67 index, etc.) of patients pathologically confirmed with or excluded from breast cancer in our center between January 2019 and December 2024. For biopsy specimens from patients diagnosed with breast cancer and immunohistochemically confirmed as HR+/Her-2+ during the same period, additional immunohistochemical staining for CD8A, KPT5, GFRA1, and PFKP should be performed, with images and results collected. The collected basic clinical information, imaging data, pathological findings, and laboratory metrics of patients will serve as candidate inputs. Units of measurement will be standardized, and missing data will be imputed using the multiple imputation by chained equations algorithm. Data harmonization will employ the Box-Cox algorithm, while min-max scaling will be used for standardization. The adaptive synthetic sampling method with a balance ratio of 0.5 will address data imbalance. For the collected patient data, deep learning will be applied to screen features from the images, combined with clinical significance to identify malignant risk factors. A neural network classifier will be trained on the training set data, with independent variables including breast MRI/ultrasound images, CA199, CA153, CA125, AFP/CEA, etc., and dependent variables including breast cancer status and subtype. Pathological biopsy results will be set as the validation standard. Model tuning will be conducted on the validation set to construct a breast cancer prediction model. It should be noted that as a single-center study, the results have limited generalizability. The further optimization and evaluation plan for the model involves using breast disease screening data from external centers for validation and refinement, evaluating the model's practical impact on clinical decision-making, and continuously tracking and optimizing its performance.

The goal of this observational study is to evaluate the diagnostic accuracy of different deep learning models in detecting dental caries from intra oral images taken by a professional intra oral camera in children. The main question it aims to answer is: What is the diagnostic accuracy of different deep learning models in detecting dental caries from intra oral images taken by a professional intra oral camera in children compared to the conventional clinical visual examination?

The objective of this prospective observational study is to rigorously examine the feasibility and efficacy of utilizing latent diffusion models for data augmentation in anti-nuclear antibody (ANA) Hep-2 cell immunofluorescence images. The main question it aims to answer is: Can the application of such models potentially enhance the data quality, increase sample diversity, or improve the accuracy and efficiency of subsequent analytical processes (like disease diagnosis and classification) when utilized with ANA-related images?