The study, published as an open-access research article in BMC Medical Imaging, was led by Lanlan Li, Yu Zeng and Ziyue Wang—joint first authors—along with colleagues at Fuzhou University, Sun Yat-sen University’s Sixth Affiliated Hospital in Guangzhou, and Guangdong Second Provincial General Hospital. Their central insight is deceptively simple: if real clinical images are scarce, why not synthesize realistic, diverse, diagnostically relevant ones? The difficulty, as any practitioner of medical image generation knows, lies in making synthetic images that are simultaneously faithful to the underlying pathology and varied enough to genuinely help a classifier learn.

Colorectal cancer, often abbreviated CRC, is routinely screened through tools such as the fecal immunochemical test and endoscopic examination. In parallel, tongue diagnosis—a pillar of traditional East Asian medicine—has attracted growing scientific interest because changes in tongue color, coating, texture and shape can correlate with systemic disease states. Deep-learning classifiers trained on tongue photographs have shown promise in distinguishing CRC patients from healthy controls. But these models are voracious consumers of data, and clinical tongue image collections are typically small, imbalanced and expensive to curate. Conventional augmentation techniques—rotations, flips, brightness shifts, crops—merely rearrange existing pixels without creating new pathological variation, and they cannot mimic the complex, clinically meaningful differences that separate diseased from healthy tongues.

The research team’s answer was to build an improved diffusion-based generative model. Diffusion models, which learn to create images by reversing a gradual noising process, have taken the computer vision world by storm in recent years. The researchers started from the Denoising Diffusion Implicit Model, or DDIM, a fast-sampling variant of the widely used DDPM family, and then substantially re-engineered its denoising backbone, a U-Net neural network, in two important ways.

The first innovation is the embedding of a DeepSeek Mixture-of-Experts, or DeepSeekMoE, module into the denoising U-Net. Mixture-of-experts architectures are a class of sparse neural networks in which a gating mechanism routes each input to specialized subnetworks—”experts”—rather than pushing all data through a single monolithic set of weights. In DTMG-Net, this routing operates adaptively across the diffusion timesteps: early in the reverse process, when an image is nearly pure noise, the model faces a very different task than in later steps, when fine anatomical and textural details must be resolved. By allowing different expert subnetworks to specialize at different stages of denoising and at different scales of pathological features, the model gains a more nuanced capacity to represent the multi-scale structure of tongue imagery, from the global shape of the tongue body down to the subtle lesions that matter diagnostically.

The second innovation is a multi-scale dilated attention block, or MSDA, placed at the bottleneck of the U-Net—the point in the network where spatial resolution is lowest and semantic abstraction is highest. Dilated attention applies attention mechanisms across receptive fields that have been expanded with dilation, letting the network capture both long-range, global structure (the overall geometry and color distribution of the tongue) and fine-grained local texture (coating patterns, cracks, and lesion-like features) without the prohibitive memory cost of full-resolution attention. The joint capture of global form and local detail is precisely what makes a synthetic tongue image look plausible both at a glance and under the scrutiny of a trained classifier.

A third element of the design targets a well-known failure mode of generative models: mode collapse, in which a generator learns to produce a narrow set of safe, repetitive outputs rather than exploring the full diversity of the data distribution. The team designed a joint loss function that combines the standard noise-estimation objective of diffusion training with an intra-sample diversity regularization term. In effect, the model is rewarded not only for reconstructing realistic images but also for producing variations that differ meaningfully from one another, expanding the visual richness of the synthetic tongue samples without sacrificing fidelity.

Quantitatively, the authors evaluated their generated images using two standard metrics in generative modeling. The Fréchet Inception Distance, or FID, measures the statistical distance between the distribution of generated images and that of real images—lower is better. The Inception Score, or IS, rewards generators for producing images that are both classifiable and diverse—higher is better. Benchmarked against a variational autoencoder (VAE), a deep convolutional generative adversarial network (DCGAN), the PNDM diffusion model, and the original DDIM, DTMG-Net achieved FID values of 73.83 for CRC tongue samples and 58.99 for healthy control samples. The researchers are candid that these absolute FID values remain relatively high—a reflection of the exceptionally small-sample setting of tongue image data, which makes distribution matching inherently difficult. What matters, they argue, is that DTMG-Net attained the smallest distribution discrepancy of any generative model compared in the study under these challenging conditions.

Ablation experiments—experiments in which individual components are removed to test their contribution—confirmed that both the DeepSeekMoE module and the MSDA block independently improved generation performance, and that the diversity constraint raised the Inception Score without a meaningful degradation in FID. That balance is the whole point: a generator that is realistic but repetitive teaches a classifier little; one that is diverse but implausible can actively mislead it.

The decisive test, however, was downstream. The team augmented training sets with DTMG-Net synthetic images and trained four mainstream classification backbones: WideResNet, ResNet50, MedMamba and the Vision Transformer, or ViT. These architectures span the modern deep-learning landscape, from convolutional workhorses to state-space hybrids and transformer-based models. Across these backbones, training sets enriched with DTMG-Net generated images generally produced higher numerical values of the area under the ROC curve (AUC), F1-score, and accuracy than training sets augmented with images from the competing generative strategies. The improvement was consistent in direction though, as the authors note, “numerical” in character—measured within the constraints of their available dataset—yet the pattern held across diverse architectures, which strengthens the case that the synthetic data carries genuine diagnostic signal rather than dataset-specific noise.

The clinical implications are noteworthy. Screening for colorectal cancer remains imperfect: adherence to colonoscopy is limited, and non-invasive tests such as the fecal immunochemical test have well-documented sensitivity constraints. A supplementary, entirely non-invasive modality based on an ordinary photograph of the tongue—an examination that costs nothing, causes no discomfort, and requires no laboratory infrastructure—could, if validated at scale, complement existing screening pathways, particularly in resource-limited settings or in large-scale community screening campaigns where endoscopic capacity is scarce.

The study also exemplifies a broader and rapidly growing trend in medical AI: the use of generative augmentation to rescue small clinical datasets. Rare diseases, niche imaging modalities and traditional medicine modalities alike suffer from data scarcity that prevents deep networks from reaching their potential. Generative approaches of the kind embodied in DTMG-Net offer a path forward that conventional augmentation cannot: the creation of new, plausible, pathological variation rather than mere geometric transformations of existing images. The researchers emphasize that their scheme requires no complex preprocessing, which lowers the barrier to deployment in practical clinical workflows.

The work was supported by the General Program of the National Natural Science Foundation of China, and was conducted under ethical approval from the Medical Ethics Committee of the Sixth Affiliated Hospital of Sun Yat-sen University, with informed consent from all subjects and anonymized research data. The article was published open access on 8 September 2026, having been received on 30 June and accepted on 31 August of that year, and is available under a Creative Commons license.

Caveats remain, as they always do at this stage of translational research. FID values in the tens indicate that synthetic tongue images are still not statistically indistinguishable from real ones, and downstream performance gains, while consistent, were evaluated on small-scale datasets. Prospective, multi-center validation with large, diverse patient cohorts will be needed before any tongue-image-based CRC screening tool reaches the clinic. Privacy and data-governance considerations around synthetic medical imagery will also demand careful attention.

Nevertheless, DTMG-Net offers a compelling demonstration that ideas flowing from the frontiers of generative AI—sparse mixture-of-experts routing, dilated multi-scale attention, diversity-aware training objectives—can be purpose-built for problems as specific and as human as reading a tongue. In doing so, the study points toward a future in which the ancient diagnostic art of tongue inspection is augmented, rather than replaced, by machines trained to see patterns that even experienced clinicians might miss. It also adds to the accumulating evidence that generative augmentation is becoming an indispensable tool in the medical AI toolkit, turning small, hard-won clinical datasets into the training corpora that modern deep networks demand.

The event, taking place from July 15 to 19 at the Taipei Nangang Exhibition Center, will gather over 850 exhibitors from nearly 60 countries, emphasizing cutting-edge biopharmaceutical research, AI-driven healthcare solutions, and cross-border digital health innovations. Dr. Zhavoronkov’s participation in a panel discussion on July 16, focused on AI × Medicine: Reshaping the Future of Drug Discovery, further underscores his role as a thought leader in this transformative space.

Insilico Medicine has pioneered the integration of generative AI and automation in drug discovery, drastically compressing timelines from target identification to the nomination of development candidates. This accelerated pipeline not only enhances scientific productivity but also introduces a self-improving AI platform that benefits from continuous learning across diverse research programs. Dr. Zhavoronkov’s keynote will explore these core pillars of sustainability, spotlighting how rigorous benchmark-driven productivity, strategic portfolio management, and AI-powered innovation coalesce to build a robust biotech enterprise.

The talk also promises technical insights into the novel AI frameworks Insilico employs to target complex diseases such as fibrosis, oncology, immunology, and metabolic disorders. By leveraging deep learning algorithms and automation, Insilico’s platform optimizes molecular design and candidate screening, enabling rapid iteration and refinement that conventional methods cannot match.

BIO Asia-Taiwan 2026, themed “Asian Inspiration, Global Impact,” provides a dynamic platform fostering international collaboration among life science leaders, investors, and innovators. The event’s integration of business partnering sessions and exhibitions aims to accelerate the translation of AI-driven discoveries into tangible healthcare solutions, reinforcing the pivotal role of technology in future drug development.

Insilico Medicine, publicly listed on the Hong Kong Stock Exchange since December 2025 (HKEX:3696), exemplifies the next-generation biotech company. Their approach extends beyond pharmaceuticals, applying Pharma.AI technologies to sectors like advanced materials, agriculture, and veterinary medicine, broadening the impact of AI innovations across multiple industries.

As AI continues to disrupt traditional drug discovery paradigms, events like BIO Asia-Taiwan become crucial forums for knowledge exchange and partnership building. Dr. Zhavoronkov’s involvement highlights the growing importance of sustainability-driven biotech strategies powered by AI, signaling a promising future for longevity-focused therapeutics and beyond.

This convergence of AI and life sciences not only accelerates the pace of innovation but also redefines how companies sustain growth, adapt, and continually generate value in the fast-evolving biotech landscape.

Subject of Research: AI-driven drug discovery and sustainable longevity biotech companies
Article Title: Spotlighting Longevity and AI Drug Discovery: Insilico Medicine to Showcase at BIO Asia-Taiwan 2026
News Publication Date: Not specified
Web References: www.insilico.com
Image Credits: Insilico Medicine
Keywords: Longevity, Artificial Intelligence, Drug Discovery, Generative AI, Biotechnology, Automation, Pharma.AI, BIO Asia-Taiwan

Dragon Copilot is architected on a highly secure, scalable, and healthcare-specific foundation that merges advanced natural language processing, ambient listening technology, and generative AI capabilities. These sophisticated technologies enable seamless documentation by capturing physician-patient interactions in real time, directly integrating comprehensive clinical notes into the electronic health record (EHR) system. By automating cumbersome administrative duties and surfacing critical clinical data instantaneously, Dragon Copilot facilitates a smoother, more efficient clinical workflow and drastically reduces the time clinicians spend on paperwork.

Lisa Stump, Executive Vice President and Chief Digital Information Officer at Mount Sinai Health System, emphasized that after a rigorous multi-vendor evaluation, this deployment of Dragon Copilot signifies a transformative leap in technological support for clinicians. The AI assistant is directly embedded within the clinical workflow, effectively alleviating the pervasive issue of documentation fatigue while enhancing information flow. This, in turn, frees healthcare providers to dedicate more time and attention to patient care—ultimately advancing Mount Sinai’s mission of delivering exceptional health services.

The underlying technology employs unified voice recognition and ambient listening, allowing caregivers—including physicians and nurses—to document care through natural, conversational dialogue during clinical encounters. These capabilities are augmented by generative AI, which synthesizes and organizes the documented information, simplifying intricate clinical tasks and improving care coordination. This results in higher documentation accuracy and speed, fostering richer interactions with patients who benefit from more attentive and engaged providers.

Brendan G. Carr, MD, MA, MS, CEO of Mount Sinai Health System, articulated the institution’s responsible approach to AI integration, underscoring that Dragon Copilot is intended not to supplant human clinical judgment but to empower it. The AI acts as an extension of the care team, enhancing human decision-making by providing timely, contextual information and streamlining workflows. This partnership between human expertise and artificial intelligence is designed to facilitate smarter, more interconnected, and compassionate care delivery.

Kenneth Harper, General Manager of the Dragon product for Microsoft Health and Life Sciences, acknowledged Mount Sinai’s leadership in academic medicine innovation. He highlighted that embedding ambient and generative AI directly into clinical processes helps reduce clinician burnout and reclaim valuable time, enabling healthcare professionals to reconnect with their patients meaningfully. This deployment exemplifies a shared vision of responsible AI use that centers the human element within the healthcare experience.

The phased rollout of Dragon Copilot has commenced with select departments at Mount Sinai, with plans for a comprehensive system-wide implementation scheduled for 2026. Each stage includes meticulous training programs, ongoing feedback mechanisms, and detailed evaluations to ensure a secure, equitable, and effective integration. This careful, iterative approach helps safeguard both clinician satisfaction and patient safety while optimizing AI tool adoption.

Mount Sinai’s commitment to AI-driven innovation is part of a broader digital strategy aimed at enhancing clinician well-being, operational efficiency, and clinical outcomes. By adopting sophisticated AI companions like Dragon Copilot, the health system reinforces its status as a national exemplar of innovation in academic medicine, poised to transform healthcare delivery on multiple fronts.

In deploying Dragon Copilot, Mount Sinai not only reduces the cognitive burden associated with exhaustive documentation but also addresses clinician burnout—a critical concern in modern healthcare environments. The ambient AI listens passively in the background, capturing clinical narratives without disrupting the natural physician-patient dialogue. This ambient data collection is then processed by advanced generative models that produce accurate, comprehensive clinical notes and prompt necessary administrative actions, facilitating better clinician focus and improved care continuity.

Furthermore, the integration of AI-driven clinical documentation significantly enhances data interoperability and information retrieval within the EHR. The system’s capability to distill and elevate the most pertinent clinical facts ensures that providers have ready access to critical patient insight, supporting informed decision-making and enabling personalized care plans grounded in comprehensive, up-to-date information.

Mount Sinai Health System’s expansive infrastructure, consisting of seven hospitals, hundreds of outpatient practices, and numerous research centers, provides a demanding environment that benefits profoundly from such AI-enabled transformation. The scale and complexity of its clinical operations underscore the necessity for innovative solutions that can reduce the administrative burden on healthcare workers and streamline patient management across diverse care settings.

The integration of Dragon Copilot at Mount Sinai serves as a beacon for the broader medical community, showcasing how AI can be ethically and effectively harnessed to augment clinical practice. As healthcare continues to grapple with increasing data volumes, regulatory pressures, and workforce challenges, such AI assistants offer vital tools to sustain high-quality care delivery and promote clinician resilience.

For those interested in following Mount Sinai’s ongoing AI advancements, the institution maintains a dedicated resource hub, chronicling their progress in embedding artificial intelligence into healthcare systems. This provides a wealth of information about how cutting-edge AI technologies like Dragon Copilot are reshaping academic medicine and clinical care on a broader scale.

Subject of Research: Artificial Intelligence in Clinical Documentation and Workflow Automation
Article Title: Mount Sinai Health System Partners with Microsoft to Deploy AI-Powered Dragon Copilot Clinical Assistant
News Publication Date: Not explicitly provided in the text
Web References: https://icahn.mssm.edu/about/artificial-intelligence; https://www.mountsinai.org
Keywords: Artificial Intelligence, Clinical Documentation, Electronic Health Record, Generative AI, Ambient Listening, Healthcare Innovation, Clinical Workflow Automation, Digital Transformation, Clinician Burnout, Patient Care, Academic Medicine, AI in Healthcare

At the heart of this study is the acknowledgment that generative AI can enhance nursing practice in several ways. From automating administrative tasks to providing clinical decision support, AI can alleviate some of the burdens that nurses face daily. The researchers sought to identify the factors that motivate or inhibit nurses’ willingness to embrace these new technologies, contributing to a more nuanced understanding of the human-AI interface in healthcare.

As generative AI continues to make significant strides, the nursing profession finds itself at a crossroads. On one hand, nurses are excited about the potential improvements in their workflow; on the other, concerns about job displacement and ethical implications loom large. The findings of the study suggest that intrinsic motivation, training, and organizational support play crucial roles in shaping nurses’ attitudes toward AI integration. The research highlights that without addressing these factors, the widespread adoption of generative AI in nursing may face formidable obstacles.

The study surveyed a diverse group of nurses across various specialties and experience levels. By using a well-defined set of metrics to assess the factors associated with their intention to use generative AI, the research provides valuable insights into the perceptions of nurses in relation to technology within healthcare settings. One significant finding is that familiarity with AI systems correlated positively with a willingness to engage with these technologies. This underscores the necessity for comprehensive training programs that enhance nurses’ understanding and skills related to generative AI.

Moreover, the study identified that organizational culture plays an influential role in the acceptance and integration of AI in nursing practices. Environments that foster collaboration, innovation, and continuous learning are more likely to see enthusiastic adoption of these advanced technologies among their nursing staff. Organizations that prioritize technology adoption must also create robust communication strategies to clarify the benefits and usage of AI tools. Addressing concerns and encouraging dialogue can build a more favorable climate for embracing AI.

Another striking outcome of the research revealed that educational background and ongoing professional development significantly impact nurses’ intention to utilize generative AI. Those with higher levels of education expressed a greater inclination toward adopting new technologies, indicating that advanced training could be instrumental in shaping a skilled workforce prepared to leverage AI efficiently. Integrating AI into nursing curricula is a proactive step toward ensuring that upcoming generations of nurses are well-equipped to thrive in a tech-savvy healthcare landscape.

The study also spotlighted the psychological aspects influencing nurses’ perceptions of AI. Fear of job loss and doubts about AI’s reliability can deter nursing staff from utilizing generative AI. This highlights the importance of demonstrating the complementary nature of AI tools, wherein technology enhances nursing capabilities rather than replacing human expertise. Developing trust in AI as a partner in patient care will require consistent, positive experiences and transparent outcomes showcasing AI’s effectiveness.

Choe and Woo’s study emphasizes the necessity of a multi-faceted approach to encourage acceptance of generative AI among nurses. This includes engaging with nurses through feedback loops to understand their concerns and incorporating their input in the development of AI tools tailored to their needs and workflows. When nurses feel heard and involved in the process, their acceptance and willingness to embrace the changes brought by AI technologies significantly improve.

As the study indicates, the timeline for widespread implementation of generative AI in nursing is contingent upon various factors. Organizational readiness, the political landscape, and regulatory frameworks will all contribute to how swiftly these technologies can be integrated into practice. Policymakers play a critical role in promoting standards and guidelines to ensure that AI’s usage in healthcare is safe, ethical, and beneficial for both practitioners and patients.

In conclusion, Choe and Woo’s study sheds light on the pivotal factors influencing the willingness of nurses to adopt generative AI technologies. It presents a roadmap for overcoming barriers through targeted training, fostering a supportive organizational culture, and addressing ethical considerations. The healthcare industry stands at the brink of a revolution, with opportunities to transform nursing practice through innovative technology. By prioritizing education and collaboration, the nursing profession can ensure that it advances alongside technological developments, ultimately leading to enhanced patient care and efficient workflow management.

In an ever-changing healthcare environment, the focus on the human element amid technological innovations remains crucial. As generative AI becomes increasingly prevalent, understanding the perspectives of healthcare professionals will provide invaluable insights into creating a brighter future for both nursing practice and patient outcomes.

Subject of Research: Factors associated with intention to use generative artificial intelligence in nursing practice.

Article Title: Factors associated with intention to use generative artificial intelligence in nursing practice: a cross-sectional study.

Article References:

Choe, J., Woo, K. Factors associated with intention to use generative artificial intelligence in nursing practice: a cross-sectional study.BMC Nurs 24, 1327 (2025). https://doi.org/10.1186/s12912-025-03985-y

Image Credits: AI Generated

DOI: 10.1186/s12912-025-03985-y

Keywords: Generative AI, nursing practice, healthcare technology, perception of AI, organizational culture, education in nursing, adoption barriers, nursing workforce.

In a world where the Internet and AI are increasingly intertwined in our daily lives, the research sheds light on a pressing need for trustworthy health information, especially for vulnerable populations like pregnant women. As technology advances, more individuals are turning to online resources for health advice, amplifying the excellence of this research. The stakes are incredibly high in areas like opioid use during pregnancy, where delayed action or misinformation can lead to dire health consequences, not just for the expectant mother, but also for the unborn child.

Drew Herbert, the lead author of the study from the University of Missouri’s Sinclair School of Nursing, remarks on the urgency of reliable health advice, especially in the context of opioid dependence during pregnancy. “A sense of urgency surrounds this area of medical care; inaccurate information or delays in treatment can be harmful,” he states. As many health discussions often carry an element of stigma, the ability to seek information privately through AI platforms may empower individuals without the fear of judgment.

To conduct this research, the team utilized a detailed persona named “Jade,” a hypothetical pregnant woman experiencing challenges with opioid addiction. This persona was instrumental in structuring conversations that would reflect real-world queries faced by similar individuals. The researchers fed the AI model prompts that involved clinical scenarios, asking direct questions about treatment options and the process of finding a healthcare provider. This method successfully integrated personal context, demonstrating how generative AI could simulate a clinician with genuine empathy and understanding.

The researchers recorded and analyzed 30 distinct conversations with ChatGPT, evaluating its responses against a rigorous rubric designed for medical accuracy and safety. Astonishingly, the results revealed that nearly 97% of the generated responses were not only safe but also aligned with established medical protocols. ChatGPT provided information about potential treatments, including medication-assisted therapies, and guidance on finding healthcare professionals, showcasing the technology’s potential in bridging gaps within the healthcare communication landscape.

One of the study’s more surprising findings was the consistency of the AI’s responses, which adhered closely to clinically accepted practices. Herbert acknowledges this success by stating, “The level of accuracy far exceeded our initial expectations.” However, the authors also exercise caution. They recognize that the efficacy of AI-generated information is largely dependent on the specificity of the prompts provided, indicating that generic queries may yield less reliable results.

Looking forward, the researchers emphasize that this study is less about creating entirely new healthcare solutions and more about leveraging existing technology to improve patient outcomes. Fine-tuning the AI model represents a significant next step, necessitating further studies to understand how AI can be safely implemented in real-world settings. Testing in field-based environments is essential to ascertain the effectiveness of generative AI as a tool for health advice.

Further exploration is required to ensure that the AI helps users navigate their healthcare pathways while reinforcing the importance of professional medical consultations. The goal is to harness innovative technology in a way that complements existing care systems, making it easier for individuals to seek assistance without the burden of stigma.

Overall, this study stands as a testament to the transformative potential inherent in AI technology when it comes to healthcare. It emphasizes the capability of generative models to deliver safe, reliable, and clinically relevant information that can empower individuals to take control of their health challenges confidently. As society continues to grapple with the complexities of opioid addiction and its impacts on pregnant women, the findings of this research may serve as a beacon for advancing communication, understanding, and treatment options in the field.

The ongoing evolution of AI holds promising prospects for mental health, substance use disorders, and various other medical fields, suggesting that as generative technology matures, it could become a staple in the healthcare environment. The integration of such systems in people’s lives would reshape the way they access, understand, and engage with health information in an increasingly complex medical landscape.

As we stand at the precipice of this technological revolution in healthcare, it becomes imperative to not only harness the power of these tools but to do so in a responsible manner. With ongoing iterations and deeper evaluations of AI capacity, we may find ourselves entering a new era of medical care that prioritizes safety, accessibility, and the informed autonomy of every individual seeking treatment for complex health conditions.

Subject of Research: People
Article Title: Generative AI-derived information about opioid use disorder treatment during pregnancy: An exploratory evaluation of GPT-4’s steerability for provision of trustworthy person-centered information
News Publication Date: 23-Oct-2025
Web References: http://dx.doi.org/10.15288/jsad.24-00319
References: Herbert, D., Westendorf, J., Farmer, M., & Reeder, B. (2025). Generative AI-derived information about opioid use disorder treatment during pregnancy: An exploratory evaluation of GPT-4’s steerability for provision of trustworthy person-centered information. Journal of Studies on Alcohol and Drugs, 86(6), 894–905.
Image Credits: Credit: Journal of Studies on Alcohol and Drugs

Keywords

Generative AI, opioid use disorder, pregnancy, healthcare technology, ChatGPT, medical advice, clinical practice, addiction treatment, online health information, stigma, patient empowerment, health communication.

The AI model owes its innovative design to sophisticated algorithmic principles borrowed from the architecture of large language models (LLMs). Researchers harnessed anonymized health data from a substantial cohort of 400,000 patients associated with the UK Biobank, employing state-of-the-art computational methods to ensure the model’s efficacy. Despite the localized focus on UK patient data, the model demonstrated its utility by successfully forecasting health outcomes when tested against an even larger dataset of 1.9 million patients from the Danish National Patient Registry.

What sets this research apart is the holistic methodology employed, making it one of the most comprehensive undertakings in both generative AI and health risk prediction. The model meticulously learns the “grammar” of health events by treating medical histories as sequences of time-bound incidents. It recognizes the integral patterns that govern human health, including crucial lifestyle factors such as smoking or the occurrence of various medical diagnoses over an individual’s lifetime. By understanding these patterns, the AI can generate insightful forecasts about potential future health outcomes that could empower both individuals and healthcare professionals alike.

Ewan Birney, the Interim Executive Director of the European Molecular Biology Laboratory (EMBL), shared his enthusiasm regarding the AI’s transformative potential. He emphasized that the model serves as a proof of concept, illustrating the feasibility of employing AI to discern long-term health patterns. As medical knowledge continues to evolve, utilizing predictive tools could facilitate early interventions tailored to individual needs, steering the healthcare sector towards a more personalized and preventive approach.

The collaboration between EMBL, the German Cancer Research Centre (DKFZ), and the University of Copenhagen signifies a monumental step taken in understanding how illnesses evolve over time. Drawing comparisons to how large language models decode the structure of sentences, this AI model employs a similar approach to understanding health data dynamics. It finds significant correlations between medical events and aids in projecting prospective health risks. While the results are not definitive predictions, they provide valuable projections based on individual medical histories and various risk factors.

The AI model boasts a particularly impressive performance in predicting conditions that follow clear and consistent patterns, such as certain cancers, heart disease, and sepsis. The scientific community finds great value in the model’s ability to effectively forecast outcomes in these scenarios. Conversely, the model grapples with considerable challenges when addressing health conditions characterized by high variability, including mental health disorders that hinge on unpredictable life developments. Such nuances illustrate the model’s current limitations while laying the foundation for its ongoing evolution.

Although promising, the model operates on a principle similar to weather forecasting. It generates probabilities of health events rather than certainties. For instance, the AI can estimate an individual’s risk of developing heart disease within a particular timeframe, akin to predicting a 70% chance of rain the next day. The model’s efficacy diminishes in long-range forecasts due to inherent uncertainties common in all predictive models.

A closer examination of the heart attack forecasts derived from UK Biobank data reveals fascinating insights. For adult men aged 60-65, the risk of a heart attack varies significantly, with some cases presenting a one in ten thousand annual risk, whereas others may face a staggering one in one hundred odds. The model also highlights how risk escalates with age, aligning closely with observed case data, affirming its reliability in predicting health outcomes across different demographics.

However, one must emphasize that the model’s training dataset is not entirely inclusive. Predominantly comprising participants aged 40-60, the model exhibits a notable gap in addressing childhood or adolescent health events. Additionally, the dataset reflects a demographic bias that can skew risk assessments, particularly for underrepresented ethnic groups. Thus, as the field advances, rectifying these biases through more diverse datasets will be essential for enhancing the model’s applicability and fairness.

In its current form, while the model is not yet tailored for clinical application, its potential usefulness is undeniable. Researchers could leverage it to deepen their comprehension of how diseases unfold and advance over time. Moreover, the model can facilitate exploration into the impacts of lifestyle choices and previous health issues on long-term risks. It also opens avenues for health outcome simulations using artificially constructed patient data, especially in scenarios where access to real-world datasets remains a challenge.

Anticipating the future, it is evident that AI applications similar to this model, when integrated with more representative health datasets, could transform clinical practices. With aging populations and increasing chronic disease incidence, accurate forecasting of health needs would enable healthcare systems to optimize resource allocation effectively. Nevertheless, rigorous testing and the establishment of robust regulatory frameworks are pivotal before any AI-driven approach can become commonplace in clinical environments.

Moritz Gerstung, the Head of the Division of AI in Oncology at DKFZ, emphasized that this research marks the commencement of a new era in understanding human health and disease progression. The generative AI model developed here could pave the way for personalized healthcare approaches that anticipate future needs at scale. By drawing lessons from extensive populations, it offers a compelling perspective on disease development, fostering a landscape where earlier, more tailored interventions could be realized.

Importantly, the development of this AI model adhered to stringent ethical guidelines governing the use of health data. The anonymized patient information utilized from the UK Biobank was collected under informed consent, ensuring that participant privacy was paramount throughout the research process. Compliance with national regulations concerning Danish data further underscores the commitment to ethical standards in research. Secure virtual systems used for data analysis assured that sensitive information remained protected, thereby aligning the model’s development with emerging ethical mandates.

The profound implications of this generative AI model extend far beyond mere predictions. They embody the potential to revolutionize our approach to healthcare by fostering a culture of estimated risk awareness and proactive health management. Built on a foundation of rigorous science and ethical practice, this model stands poised to change the trajectory of how healthcare systems function, addressing challenges faced in disease prevention and paving the way for more informed patient care.

Subject of Research: AI and Health Risk Prediction
Article Title: Learning the natural history of human disease with generative transformers
News Publication Date: 17-Sep-2025
Web References: http://dx.doi.org/10.1038/s41586-025-09529-3
References: Nature, EMBL-EBI
Image Credits: Karen Arnott/EMBL-EBI

Keywords

Artificial intelligence, Computer modeling, Health and medicine, Clinical medicine, Diseases and disorders, Health care, Human health

Throughout history, the challenges of primary care have remained largely unchanged—long wait times, limited access to providers, and inefficiencies in healthcare delivery. The introduction of generative AI offers a potential solution, streamlining various aspects of healthcare to enhance efficiency and quality. This innovative technology acts as a foundation model, providing a framework that can adapt to diverse medical queries and patient needs.

One of the most significant benefits of generative AI is its ability to analyze vast amounts of data rapidly. In a typical primary care setting, providers often juggle multiple responsibilities, from diagnosing conditions to managing treatment plans. By incorporating AI tools, healthcare practitioners can automate routine tasks, such as appointment scheduling and data entry, thus allowing them to focus on patient interactions and clinical decision-making. The result is a more personalized approach to healthcare that truly caters to the individual needs of patients.

Moreover, generative AI has the potential to improve patient engagement significantly. Personalized communication, tailored treatment plans, and proactive follow-ups can be generated through AI algorithms that analyze patient history and preferences. This level of customization not only enhances patient satisfaction but also fosters a differential connection between patients and their providers, empowering patients to take an active role in their health journeys.

The study underscores the importance of incorporating ethical considerations into the deployment of generative AI. While the technology promises numerous advantages, there are concerns regarding data privacy and algorithmic bias that must be addressed. Researchers emphasize the need for transparency in AI systems and suggest implementing robust regulatory frameworks to safeguard patient data and ensure equitable access to care across diverse populations.

Generating insights from electronic health records (EHR) is another area where generative AI shines. Instead of healthcare providers spending hours poring over patient histories, AI can quickly synthesize relevant information, enabling doctors to make informed decisions swiftly. Such efficiency not only enhances the patient experience but can also lead to better health outcomes, as timely interventions can be initiated based on the AI-generated insights.

The implications extend beyond clinical practice, reaching into the realm of medical education and training. Generative AI can be utilized to create realistic scenarios for training future healthcare providers, exposing them to a variety of patient interactions in a controlled environment. By using simulations, medical students can enhance their diagnostic skills and learn to navigate complex patient cases, all while honing their interpersonal communication skills.

At the same time, the research team highlights challenges and limitations associated with AI in healthcare. For instance, there are inherent risks in over-reliance on AI-generated recommendations, as the nuances of human interaction and clinical intuition can sometimes be overlooked in automated processes. The blending of AI capabilities with human expertise is, therefore, crucial to ensure that patient care remains comprehensive and compassionate.

Furthermore, the introduction of generative AI necessitates a shift in training paradigms for healthcare professionals. A dual approach that combines technical training in AI technologies with traditional clinical skills will be essential. Practitioners will need to understand how to interpret AI-generated information and integrate it into their decision-making processes while maintaining strong connections with their patients.

Another key aspect of the study emphasizes the scalability of AI solutions in primary healthcare. Generative AI can cater to a broad range of healthcare settings, from urban hospitals to rural clinics, thus addressing disparities in access to care. By democratizing access to advanced healthcare tools, generative AI has the potential to bridge gaps across various populations, ultimately improving health equity.

In conclusion, the findings of Yang and his colleagues paint a compelling picture of the future of primary healthcare, where generative AI stands at the forefront. As healthcare systems increasingly embrace AI technologies, the potential for enhanced efficiency, quality, and patient engagement becomes more tangible. The journey towards a more effective primary care model is just beginning, and the role of AI will undoubtedly be integral in shaping how patients and providers interact in this new era.

As the research continues to evolve, stakeholders from across the healthcare sector must work collaboratively to harness the benefits of generative AI while remaining vigilant about its ethical implications. By prioritizing the integration of technology with human care, the future of primary healthcare can not only become more efficient but align closely with the core values of empathy and personalized care.

Subject of Research: The impact of generative AI on primary healthcare delivery.

Article Title: Transforming the Primary Care Journey with Generative AI: A Foundation Model to Boost Efficiency, Quality, and Engagement.

Article References: Yang, B., Lee, J.J., Tsai, T. et al. Transforming the Primary Care Journey with Generative AI: A Foundation Model to Boost Efficiency, Quality, and Engagement. J GEN INTERN MED (2025). https://doi.org/10.1007/s11606-025-09716-y

Image Credits: AI Generated

DOI:

Keywords: Generative AI, primary care, healthcare technology, patient engagement, efficiency, quality.

Addressing this critical bottleneck, a multidisciplinary team of researchers from IBM Research, the Cleveland Clinic, and the University of Tsukuba has unveiled a groundbreaking framework that harnesses generative AI to produce synthetic gait data. Their methodology diverges fundamentally from typical data augmentation techniques by embedding physics-based musculoskeletal simulations within the generative process. These simulations meticulously capture a spectrum of biomechanical parameters that reflect real-world heterogeneity: age-dependent musculoskeletal variations, pathological gait patterns, and the influence of different sensor configurations. By integrating this bio-physical realism with AI’s synthetic data generation capacity, the researchers have crafted a rich and diverse dataset that transcends conventional limitations and equips evolving AI models with the capacity to perform reliably across a multitude of clinical contexts.

Central to their approach is the deployment of physics-based musculoskeletal modeling, which simulates the dynamic interaction of bones, muscles, and joints during gait cycles. This mechanistic foundation ensures that generated synthetic data maintain physiological authenticity, accurately mirroring the nuances of human movement under varying health conditions. By encompassing patients as diverse as children with cerebral palsy and adults afflicted by neurodegenerative diseases, alongside healthy controls, the simulations capture a broad pathological spectrum. Moreover, by varying sensor parameters—such as camera angle and resolution in smartphone video captures—the synthetic dataset reflects real-world heterogeneity in data acquisition, enhancing the generalizability of subsequent AI models.

The team rigorously validated their framework against an extensive, real-world gait dataset comprising over 12,000 recordings from more than 1,200 individuals. This cohort included patients with cerebral palsy, Parkinson’s disease, dementia, and other neurological disorders, providing a challenging testbed for model evaluation. Results from these validation studies unveiled two transformative capabilities. First, models exclusively pretrained on synthetic data demonstrated “zero-shot” performance comparable to, or in some cases surpassing, models trained on real-world datasets. This finding is particularly remarkable considering that these AI models could estimate clinically significant gait parameters—such as gait speed, step length, and temporal step dynamics—and infer muscle activation patterns from single-camera videos, showcasing the efficacy of synthetic data in capturing biomechanical complexity.

Beyond zero-shot learning, the framework exhibited exceptional data efficiency in transfer learning scenarios. By initially pretraining AI models on large-scale synthetic datasets, followed by fine-tuning with limited real-world data, these hybrid models outperformed state-of-the-art deep learning architectures trained solely on extensive real-world datasets. This novel two-step approach not only maximizes the utility of scarce clinical data, especially for rare conditions, but also circumvents privacy-related obstacles by reducing reliance on large-scale patient data collection. The efficiency gains promise to accelerate the deployment of robust gait analysis tools in clinical settings, catalyzing personalized disease monitoring and management.

The implications of these findings extend significantly into the management of neurological disorders. Accurate quantification of gait aberrations aids clinicians in disease detection, severity assessment, and therapy evaluation. By facilitating precise, objective, and scalable gait analysis using readily accessible smartphone videos, this AI-driven approach could democratize neurological monitoring, particularly benefiting underserved populations with limited access to specialized motion analysis laboratories. Such scalable solutions hold the potential to complement clinical workflows, enabling longitudinal tracking of disease progression with minimal patient burden.

Moreover, the integration of physics-based musculoskeletal simulation with generative AI represents a paradigm shift in synthetic data utilization. Unlike traditional synthetic datasets limited to simple pattern replication, these bio-realistic synthetic gaits serve as a high-fidelity substitute for real clinical data, preserving mechanistic plausibility and inter-subject variability. This innovation paves the way not only for gait analysis but also for broader healthcare applications where data scarcity and privacy issues hinder AI development. Disease-specific synthetic data generation might soon become a cornerstone for training reliable, equitable AI systems across diverse biomedical domains.

The researchers’ interdisciplinary collaboration underscores the necessity of combining expertise in computational biomechanics, machine learning, and clinical neurology. Their framework bridges these domains effectively, creating a translational pathway from theoretical simulations to practical clinical tools. This synergy ensures that AI models are not only technically sophisticated but also clinically relevant, thus fostering trust and adoption among healthcare professionals. Future expansions of this work could include integrating additional sensor modalities, such as inertial measurement units or electromyography, further enriching synthetic datasets to emulate multifaceted patient monitoring scenarios.

While the study chiefly focuses on neurological disorders, its principles may generalize across various musculoskeletal and mobility-related conditions. For example, synthetic musculoskeletal simulation could enable early detection of orthopedic impairments or rehabilitative progress post-injury. By providing a scalable platform for data generation, this approach could transform clinical research paradigms, reducing dependency on exhaustive patient recruitment and invasive instrumentation, thereby accelerating innovation cycles.

Ethical considerations remain paramount in clinical AI development. By leveraging synthetic data, the framework ameliorates privacy-related ethical challenges inherent to patient data usage. Synthetic datasets mitigate risks of patient re-identification and comply seamlessly with data governance frameworks, facilitating broader research collaborations and multi-institutional validations. This ethical advantage adds additional impetus for adopting synthetic data-driven methodologies in sensitive healthcare domains.

Looking ahead, the research team envisions integrating their synthetic data approach with real-time gait monitoring applications powered by ubiquitous mobile devices. Such convergence could usher in an era of continuous, passive health monitoring, empowering patients and clinicians with timely biomarker feedback. As AI models mature, their deployment could expand into telemedicine, rural healthcare, and personalized rehabilitation, substantially influencing public health outcomes.

In summation, the novel framework developed by IBM Research, Cleveland Clinic, and University of Tsukuba redefines the boundaries of AI-driven clinical gait assessment. By synthesizing bio-realistic musculoskeletal gait data and validating their approach extensively on heterogeneous real-world datasets, the team has demonstrated a viable path to overcoming longstanding data diversity and privacy challenges. Their work heralds a future where equitable, precise, and generalizable AI tools enhance clinical decision-making and patient care across neurological and musculoskeletal healthcare domains.

Subject of Research: Development of synthetic musculoskeletal gait data for generalized and equitable AI-based clinical motion analysis.

Article Title: Utility of synthetic musculoskeletal gaits for generalizable healthcare applications

News Publication Date: July 4, 2025

Web References:
https://doi.org/10.1038/s41467-025-61292-1

References:
Arai, T., et al. (2025). Utility of synthetic musculoskeletal gaits for generalizable healthcare applications. Nature Communications. DOI: 10.1038/s41467-025-61292-1

Keywords:
Health care; Patient monitoring; Personalized medicine; Machine learning; Artificial intelligence; Neurological disorders; Computer simulation

In a thorough investigation involving 184 confirmed cases of malignant lung tumors, researchers meticulously analyzed the diagnostic performance of three prominent Gen-AI models: Gemini, Claude-3-opus, and GPT. The findings indicated that Gemini emerged as the frontrunner with an impressive accuracy rate surpassing 90%, closely followed by Claude-3-opus. In stark contrast, GPT demonstrated a considerably lower accuracy of only 65.22%. This variation in performance highlights the importance of selecting the appropriate AI model for specific tasks, particularly in high-stakes medical environments where accuracy is paramount.

Further statistical analysis elaborated on the significant advantages of Gemini’s performance in single-image interpretation compared to its counterparts. However, a concerning drop in accuracy to 58.51% was noted when Gemini was tasked with evaluating continuous CT slices. This decline can be attributed to the inherent complexities associated with interpreting lesion continuity and the requisite spatial relationships inherent in continuous imaging. Conversely, Claude-3-opus showcased a noteworthy resilience, maintaining relatively stable accuracy levels across varied image formats, which significantly enhances its operational viability in clinical settings.

One interesting aspect of this investigation was the introduction of clinical history into the diagnostic process, which marginally elevated Gemini’s accuracy to 68.30%. This increment suggests that while AI models can vastly enhance diagnostic efficiency, there remains a cautionary need to ensure these models do not become overly reliant on textual inputs, potentially overshadowing crucial imaging features that are vital for accurate diagnosis. The same pattern of underperformance in complex input scenarios was observed with GPT, which struggled with an accuracy of 48.91% when analyzing continuous CT slices, and only achieved 63.95% with integrated clinical history.

The overall performance of Claude-3-opus and GPT illuminated their robustness in handling various input types, particularly when faced with continuous imaging challenges where they exhibited notable stability and improved accuracy metrics. The study further demonstrated that by adopting a standardized diagnosis approach using consistent results across multiple attempts, Claude-3-opus consistently outperformed both Gemini and GPT in diagnostic accuracy.

To enhance the diversity of the analyzed sample, the research incorporated additional categories such as non-malignant nodules, inflammatory lesions, and normal lung appearances. This inclusion allowed for a more holistic evaluation of the AI models’ diagnostic capabilities, with results indicating that both Claude and Gemini achieved an area under the curve (AUC) of 0.61 when assessing single CT images. However, the complexity of the input directly correlated with a decline in both models’ diagnostic AUC as the input scenarios escalated.

In a crucial aspect of the research, the simplification of prompts utilized during the AI diagnostics significantly enriched the performance metrics for all three models. Post-simplification assessments recorded notable increases in AUC values, with Claude, Gemini, and GPT achieving AUCs of 0.69, 0.76, and 0.73, respectively. These enhanced metrics not only underline the importance of prompt precision but also highlight the intricate balance between model complexity and interpretative clarity, particularly when confronted with varying imaging modalities.

When assessing the diagnostic capacity of Gen-AI with respect to identified lesion features, it became apparent that Claude and GPT exhibited superior performance in both the accuracy and diversity of lesion localization and description. All models heavily featured morphological and margin characteristics as essential factors for malignancy, notably emphasizing identifiers such as “spiculated” and “irregular” lesions. This reliance on morphological features underscores a potential area for enhancement within AI models, particularly the need for more nuanced training that incorporates advanced lesion characteristics alongside more conventional identifiers.

The study’s findings brought to light a compelling issue concerning misdiagnosed cases, in which notable discrepancies were identified across various dimensions of model performance. The significant variances in misdiagnosis not only raise concerns about feature fabrication but also prompt questions regarding the maturity of the Gen-AI models in accurately learning and reproducing image features during training processes. This facet reveals a critical avenue for ongoing research aimed at refining the robustness of AI-driven diagnostic capabilities.

For performance optimization, advanced regression techniques were employed, resulting in AUCs of 0.896 and 0.884 prior to and following cross-validation, respectively. These promising results indicate not only the potential for enhanced diagnostic accuracy but also the stability of the model constructs utilized. Stepwise regression yielded similar AUC results, albeit with a noted increase in variability. Notably, external validation through datasets such as TCGA-LUAD, TCGA-LUSC, and MIDRC-RICORD-1A further substantiated the initial findings, demonstrating that Claude consistently outperformed its counterparts during assessments employing simplified prompts.

Through rigorous dimensionality reduction of features, Lasso regression contributed to a recalibration of performance indicators, achieving more balanced outcomes as validated by additional ROC curve analyses. This multi-faceted evaluation reinforces the imperative for continual refinement of Gen-AI technologies within radiology, highlighting areas for improvement and urging the incorporation of diverse datasets for comprehensive model training.

On a broader scale, the implications of these findings extend beyond mere diagnostic metrics; they pose critical questions about the future role of AI in personalized medicine. As Gen-AI technologies evolve, their integration into standard clinical practice necessitates a thoughtful approach, balancing enhanced diagnostic capabilities with the essential human elements of interpretation and decision-making in healthcare. The path forward must prioritize rigorous validation with real-world data to ensure that these promising technologies deliver on their potential, ultimately improving patient outcomes and fostering a more informed approach to radiological diagnostics.

As the landscape of healthcare continues to evolve, the intersection of AI and clinical practice will certainly be a focal point of ongoing research and discussion. With significant strides made in the evaluation of Gen-AI models, the future is poised for innovations that not only enhance diagnostic accuracy but also fundamentally transform the way radiologists and clinicians approach patient care in an increasingly complex diagnostic environment.

Subject of Research: Diagnostic accuracy of AI models for pulmonary CT images
Article Title: Enhancing Diagnostic Precision with Generative AI in Pulmonary Radiology
News Publication Date: TBD
Web References: TBD
References: TBD
Image Credits: THE AUTHORS

Keywords

Artificial Intelligence, Radiology, Pulmonary Imaging, Diagnostic Accuracy, Machine Learning, Gen-AI, Healthcare, Cancer Detection, CT Imaging, Data Analysis, Model Validation, Medical Technology.