Close Menu

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot

    New programmable photonic chip can control the speed at which light travels

    July 22, 2026

    Five epigenetic clocks reveal common and distinct signals of aging

    July 22, 2026

    Magic mushrooms may reshape your brain even after your trip is over

    July 22, 2026
    Facebook X (Twitter) Instagram
    Facebook X (Twitter) Instagram
    Health Magazine
    • Home
    • Environmental Health
    • Health Technology
    • Medical Research
    • Mental Health
    • Nutrition Science
    • Pharma
    • Public Health
    • Discover
      • Daily Health Tips
      • Financial Health & Stability
      • Holistic Health & Wellness
      • Mental Health
      • Nutrition & Dietary Trends
      • Professional & Personal Growth
    • Our Mission
    Health Magazine
    Home » News » AI beats primary care physicians in simulated diagnostic study using images and electrocardiograms
    Discover

    AI beats primary care physicians in simulated diagnostic study using images and electrocardiograms

    healthadminBy healthadminMay 18, 2026No Comments5 Mins Read
    AI beats primary care physicians in simulated diagnostic study using images and electrocardiograms
    Share
    Facebook Twitter Reddit Telegram Pinterest Email


    A nationally recognized medical AI system interpreted images, electrocardiograms, and clinical documents during live diagnostic chats and outperformed primary care physicians in mock consultations, while raising urgent questions about how such tools should be tested before actual treatment.

    Research: Advances in conversational diagnostic AI through multimodal reasoning. Image credit: Explode / Shutterstock

    Research: Advances in conversational diagnostic AI through multimodal reasoning. Image credit: Explode / Shutterstock

    In a recent study published in the journal natural medicineresearchers describe the development of a multimodal extension of the Articulate Medical Intelligence Explorer (AMIE). The model is designed to utilize a state-aware reasoning framework to manage clinical conversations and interpret visual artifacts. The study then conducted a randomized, blinded exploratory study that included 105 multimodal clinical scenarios, simulating 210 telemedicine consultations, and compared AMIE’s performance to that of 19 academically certified primary care physicians (PCPs).

    According to the study results, the new multimodal model outperformed PCP on 29 out of 32 evaluation axes, including consultation quality indicators such as diagnostic accuracy and empathy. These findings suggest that multimodal AI may ultimately be able to support telehealth delivery, pending real-world validation.

    Background on multimodal clinical AI

    Morbidity risks associated with delayed access to care are increasingly documented in global health care delivery, and experts attribute this pattern to increasing pressures from clinician burnout, healthcare fragmentation, and an aging global population. Generative AI shows promise in mitigating these challenges, but early healthcare large-scale language model (LLM) implementations were primarily limited to text-only chatbots.

    Field reviews highlight that this “text-only” constraint departs from standard clinical practice, where much diagnostic information is derived from history taking and physical examination, often supplemented by visual data.

    These limitations are particularly evident in telemedicine settings, where patients are reported to frequently exchange multimodal information with clinicians, such as skin photos taken with smartphones, electrocardiogram (ECG) traces, or scanned test reports.

    AMIE Multimodal Inference Research Design

    This study aimed to address this persistent medical AI limitation by developing a multimodal system that can emulate the structured reasoning of experienced clinicians by strategically requesting and interpreting these visual artifacts during live diagnostic consultations.

    The system, named “AMIE,” is built on the Gemini 2.0 Flash foundation model and powered by a new “state-aware” inferential temporal inference framework. AMIE’s custom architecture is designed to allow the model to maintain an internal “patient state” that tracks each patient’s chief complaint, current disease history, and prioritized knowledge gaps.

    During clinical use, this framework was constructed to specifically direct diagnostic consultation through three sequential phases:

    Obtain medical history. The system repeatedly updates patient profiles and identifies gaps in information. Additionally, the model determines whether and when multimodal artifacts are requested to enhance understanding of the patient’s medical history.

    Diagnosis and management. The system then generates a differential diagnosis (DDx) report that provides patient explanation and management guidance for the most relevant identified conditions.

    During follow-up, AI processes and clarifies patient concerns and communicates the final management plan, ensuring clarity for the patient or caregiver.

    Model performance was validated using a synchronous chat-adapted Objective Structured Clinical Examination (OSCE) format, and AMIE was evaluated against 19 primary care physicians (PCPs). The patient cohort consisted of 25 verified patient officials who participated in a total of 210 visits, two for each scenario.

    Test scenarios are based on real-world datasets, including the Skin Condition Image Network (SCIN) for dermatology, PTB-XL for ECG tracing, and selected clinical documents.

    Performance was assessed by 18 specialist physicians using the Multimodal Understanding and Handling (MUH) rubric, Practical Assessment of Laboratory Skills (PACES), and General Medical Council Patient Questionnaire (GMCPQ).

    Diagnostic accuracy and examination findings

    OSCE evaluation data showed that multimodal AMIE demonstrated significant performance advantages over PCP in both objective accuracy and subjective quality measures in 29 of the 32 metrics evaluated.

    When assessing diagnostic accuracy, statistical modeling confirmed that the AI’s DDx list was more accurate and comprehensive than the human physician’s list (P < 0.001). Accuracy was analyzed across lists containing 1 to 10 diagnoses, but neither AMIE nor PCP consistently submitted 10 differential diagnoses. Across all modalities, AI's top-k accuracy consistently outperformed PCP performance on lists containing 1 to 10 diagnoses.

    In another automated ablation analysis across a clinical document scenario, the AI’s top-1 accuracy reached 0.98, compared to 0.89 for the “vanilla” baseline Gemini 2.0 flash model. This shows that state-aware inference improved performance over the base model alone.

    For multimodal inference and overall robustness evaluation, expert ratings using the MUH rubric favored AI on seven out of nine metrics. AMIE was found to be particularly robust to variations in image quality, with low-quality images causing a greater reduction in the diagnostic performance of PCP than AMIE. In this simulated evaluation, AMIE also showed fewer less severe artifact-related false alarm events than PCP (P < 0.001).

    Furthermore, the patient role rated the AI ​​significantly higher on 10 out of 11 GMCPQ criteria, including showing empathy and listening. In multimodal tasks, AI was rated more favorably for its ability to explain findings (P < 0.01).

    Impact of conversational diagnostic AI

    Using data representing real-world clinical scenarios, this study highlights how state-of-the-art AI models can achieve performance that matches or exceeds that of PCPs in these simulated diagnostic settings by integrating perceptual evidence and state-aware reasoning.

    Despite these results, the researchers cautioned that the study was an exploratory study and not a randomized clinical trial. Future work should evaluate the system’s performance, safety, reliability, impact on clinical workflow, and health equity in real-world settings before considering clinical deployment.

    Click here to download your PDF copy.



    Source link

    Visited 14 times, 1 visit(s) today
    Share. Facebook Twitter Pinterest LinkedIn Telegram Reddit Email
    Previous ArticleSleep deprivation and endless video scrolling form a predictable behavioral loop
    Next Article AI eye scan reveals who may be at higher risk of osteoporosis
    healthadmin

    Related Posts

    Five epigenetic clocks reveal common and distinct signals of aging

    July 22, 2026

    Wistar scientists develop new bispecific T-cell engager to target ovarian cancer

    July 21, 2026

    Swimming may be clinically effective in reducing disability caused by chronic low back pain

    July 21, 2026

    Repeated weight gain and loss can accelerate muscle loss

    July 21, 2026

    Streamline pulmonary embolism care communication with an innovative framework

    July 21, 2026

    City of Hope launches trial to map how tumors change during immunotherapy resistance

    July 21, 2026
    Add A Comment

    Comments are closed.

    Categories

    • Daily Health Tips
    • Discover
    • Environmental Health
    • Exercise & Fitness
    • Featured
    • Featured Videos
    • Financial Health & Stability
    • Fitness
    • Fitness Updates
    • Health
    • Health Technology
    • Healthy Aging
    • Healthy Living
    • Holistic Healing
    • Holistic Health & Wellness
    • Medical Research
    • Medical Research & Insights
    • Mental Health
    • Mental Wellness
    • Natural Remedies
    • New Workouts
    • Nutrition
    • Nutrition & Dietary Trends
    • Nutrition & Superfoods
    • Nutrition Science
    • Pharma
    • Preventive Healthcare
    • Professional & Personal Growth
    • Public Health
    • Public Health & Awareness
    • Selected
    • Sleep & Recovery
    • Top Programs
    • Weight Management
    • Workouts
    Popular Posts
    • 1773313737_bacteria_-_Sebastian_Kaulitzki_46826fb7971649bfaca04a9b4cef3309-620x480.jpgHow Sino Biological ProPure™ redefines ultra-low… March 12, 2026
    • pexels-david-bartus-442116The food industry needs to act now to cut greenhouse… January 2, 2022
    • 1773729862_TagImage-3347-458389964760995353448-620x480.jpgDespite safety concerns, parents underestimate the… March 17, 2026
    • 1773209206_futuristic_techno_design_on_background_of_supercomputer_data_center_-_Image_-_Timofeev_Vladimir_M1_4.jpegMulti-agent AI systems outperform single models… March 11, 2026
    • 1774403998_image_28620e4b6b0047f7ab9154b41d739db1-620x480.jpgGait pattern helps distinguish between Lewy body… March 24, 2026
    • Leukemia-620x480.jpgBiomimetic platform powers CAR T therapy for… March 9, 2026

    Demo
    Stay In Touch
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    • YouTube
    • Vimeo
    Don't Miss

    New programmable photonic chip can control the speed at which light travels

    By healthadminJuly 22, 2026

    Researchers at Seoul National University and Seoul National University have developed a programmable photonic integrated…

    Five epigenetic clocks reveal common and distinct signals of aging

    July 22, 2026

    Magic mushrooms may reshape your brain even after your trip is over

    July 22, 2026

    Why do 24% of workers stay in jobs they hate? Health Insurance

    July 22, 2026

    Subscribe to Updates

    Get the latest creative news from SmartMag about art & design.

    HealthxMagazine
    HealthxMagazine

    At HealthX Magazine, we are dedicated to empowering entrepreneurs, doctors, chiropractors, healthcare professionals, personal trainers, executives, thought leaders, and anyone striving for optimal health.

    Our Picks

    Why do 24% of workers stay in jobs they hate? Health Insurance

    July 22, 2026

    Coffee may protect your heart, but energy drinks can be harmful

    July 22, 2026

    Prenatal THC exposure causes teenage binge drinking in female rats

    July 22, 2026
    New Comments
      Facebook X (Twitter) Instagram Pinterest
      • Home
      • Privacy Policy
      • Our Mission
      © 2026 ThemeSphere. Designed by ThemeSphere.

      Type above and press Enter to search. Press Esc to cancel.