Hasty Briefsbeta

All tags

#artificial intelligence

278 stories total

Bilingual

Development and Validation of a Computational Histology Artificial Intelligence-Powered Predictive Biomarker for Selection of Chemotherapy in Advanced Pancreatic Cancer - PubMed
3 months ago
- Development of a Computational Histology Artificial Intelligence (CHAI) platform to predict chemotherapy benefit in advanced pancreatic ductal adenocarcinoma (PDAC).
- The CHAI platform analyzed histomorphologic features from hematoxylin and eosin-stained biopsies to create a GvF biomarker.
- The biomarker dichotomized patients into G-pref (gemcitabine-based) or F-pref (fluoropyrimidine-based) treatment groups.
- Validation showed F-pref patients had better outcomes with F-chemo (TNTD: 8.6 vs. 7.5 months; OS: 14.4 vs. 11.7 months).
- G-pref patients had superior TNTD with G-chemo (9.6 vs. 7.2 months) but no OS difference.
- Propensity score-weighted analysis confirmed the biomarker's predictive value (TNTD P < .001; OS P = .005).
- RNA subtypes did not predict differential treatment effects (P = .3).
- The GvF biomarker can guide optimal first-line chemotherapy selection for advanced PDAC.
Enhancing therapeutic outcomes with artificial intelligence for HR-positive, HER2-negative metastatic breast cancer - PubMed
3 months ago
- CDK4/6 inhibitors are a first-line treatment for HR-positive metastatic breast cancer, but resistance develops.
- Fan et al. report results from the LINUX trial, a phase 2 study using AI digital pathology to guide therapy post-CDK4/6i progression.
- The study highlights the potential of AI in improving therapeutic outcomes for resistant cases.
- Conflict of interest disclosures include consulting and research support from major pharmaceutical companies for the authors.
Artificial intelligence agents in healthcare research: A scoping review - PubMed
3 months ago
- Artificial Intelligence (AI) agents are transforming healthcare delivery by enabling real-time decision support and patient interaction.
- A scoping review was conducted to analyze AI agent research in healthcare, following PRISMA-ScR guidelines.
- The review included 43 studies, with 36 published in 2025, categorized into conversational agents, workflow/automation assistants, and multimodal decision support agents.
- Core mechanisms involved external tool use for grounding and iterative self-correction for refinement.
- Evaluation settings were mostly simulated environments or laboratory studies, with few clinical pilots or real-world deployments.
- Reported outcomes focused on process measures and diagnostic accuracy, with limited attention to clinical outcomes and safety endpoints.
- Future research should prioritize clinical trials and robust assessment of safety, usability, and clinical efficacy.
Study: Self-generated Agent Skills are useless
3 months ago
- Agent Skills are structured packages of procedural knowledge that enhance LLM agents during inference.
- SkillsBench is introduced as a benchmark with 86 tasks across 11 domains, paired with curated Skills and deterministic verifiers.
- Tasks are evaluated under three conditions: no Skills, curated Skills, and self-generated Skills.
- Curated Skills improve average pass rates by 16.2 percentage points, with significant variation across domains.
- Self-generated Skills show no average benefit, indicating models cannot reliably create the procedural knowledge they benefit from.
- Focused Skills with 2–3 modules outperform comprehensive documentation.
- Smaller models with Skills can match the performance of larger models without them.
How AI is affecting productivity and jobs in Europe
3 months ago
- Europe faces a critical choice in AI adoption, with potential for productivity gains but risks of exacerbating inequality.
- Europe lags behind the US and China in AI innovation but shows similar adoption levels, with significant disparities among EU countries.
- AI adoption in Europe increases labor productivity by 4% on average, without short-term job losses, but benefits are unevenly distributed.
- Larger firms and those in financially developed EU countries adopt AI more readily, highlighting the role of resources and scale.
- Complementary investments in software, data infrastructure, and workforce training significantly amplify AI's productivity effects.
- Policy implications include supporting smaller firms, advancing financial markets, and focusing on complementary investments and skills development.
Large Language Model Reasoning Failures
2 months ago
- Large Language Models (LLMs) exhibit reasoning failures despite their advanced capabilities.
- A novel categorization framework divides reasoning into embodied and non-embodied types, with non-embodied reasoning further split into informal (intuitive) and formal (logical).
- Reasoning failures are classified into three types: fundamental failures, application-specific limitations, and robustness issues.
- The survey provides definitions, root causes, and mitigation strategies for each type of reasoning failure.
- A GitHub repository is released to compile research on LLM reasoning failures, serving as a resource for future studies.
Artificial intelligence and digital health in heart failure: advances in diagnosis, monitoring, phenotyping, and digital biomarkers - PubMed
2 months ago
- Artificial intelligence and digital health advancements in heart failure diagnosis, monitoring, phenotyping, and digital biomarkers.
- Keywords include Artificial intelligence, Digital health, Heart failure, Remote patient management, and Wearables.
- Conflict of interest disclosures include consulting/research support from various pharmaceutical companies.
- References cite studies on global heart failure burden, medical device readiness, AI-enabled screening for COVID-19, and AI in ECG screening for cardiac dysfunction.
Transforming nephrology through artificial intelligence: a state-of-the-art roadmap for clinical integration - PubMed
2 months ago
- AI is set to transform nephrology by enabling early detection, precise risk stratification, and clinical decision support.
- Applications of AI in nephrology include acute kidney injury (AKI), chronic kidney disease (CKD), dialysis, and kidney transplantation.
- In AKI, predictive models show strong performance but face challenges in clinical workflow integration.
- In CKD, machine learning aids in risk stratification and personalized therapy planning.
- AI enhances dialysis management by optimizing ultrafiltration, anemia control, and vascular access surveillance.
- Generative AI and large language models improve clinical documentation, triage, and patient education.
- In transplantation, AI assists in organ allocation, graft monitoring, and rejection classification.
- Challenges include data heterogeneity, bias, interpretability, regulatory uncertainty, and workflow integration.
- Future directions include multimodal data integration, reinforcement learning, digital twins, and ambient intelligence.
- Regulatory frameworks are evolving, emphasizing clinician involvement in AI model development and deployment.
Research priorities for data science and artificial intelligence in global health: an international consensus exercise - PubMed
2 months ago
- Research priorities for data science and AI in global health were identified through an international consensus exercise.
- 155 research ideas were scored by 51 experts based on feasibility, impact, paradigm shift potential, implementation, and equity.
- Top priorities included AI-based outbreak prediction, improved diagnostics for infectious diseases, and early-warning systems.
- Other high-ranking topics were AI-assisted resource allocation, telemedicine, mobile health services, and chronic disease management tools.
- LMIC experts prioritized infectious disease control and diagnostic equity, while high-income country experts focused on infrastructure and climate-related analytics.
- The study aims to align AI and data science research with global health priorities, especially in LMICs.
Artificial intelligence-generated marathon training programs: reliable tools in exercise prescription for athletic performance? - PubMed
2 months ago
- Marathon running is a global phenomenon with increasing participation across various age and experience groups.
- Training for a marathon requires adherence to principles like pacing, training volume, and periodization.
- AI-generated marathon training programs were evaluated for reliability in exercise prescription.
- AI models like Claude 3.5, ChatGPT, Gemini, and DeepSeek were used to generate 6-month training plans for different athlete levels.
- Most AI systems correctly identified key training components such as weekly mileage progression, tapering, and intensity distribution.
- Variations in AI responses included omissions of key details, failure to differentiate athlete levels, and inconsistent pacing data.
- AI shows potential in providing accessible and structured training content but lacks personalization and professional oversight.
- Future research should focus on real-world outcomes of AI-generated programs and integration of physiological data.
Artificial intelligence-generated synthetic data for cancer research and clinical trials - PubMed
2 months ago
- Artificial intelligence-generated synthetic data is increasingly used in healthcare research, especially in oncology and haematology.
- Synthetic data replicates real-world data properties, aiding in overcoming data access barriers and reducing clinical trial costs and failure rates.
- Challenges include lack of standardization in training data selection, model evaluation, bias mitigation, privacy preservation, and quality assurance.
- Synthetic data has potential to transform data sharing, scientific collaboration, and clinical trial design with rigorous validation and oversight.
- The article reviews real-world examples, limitations, and proposes best practices for enhancing synthetic data fidelity, validity, fairness, and utility.
Can AI write reports like a radiologist? A blinded evaluation of large language model-generated lumbar spine MRI reports - PubMed
2 months ago
- Study compares quality of LLM-generated vs. radiologist-written lumbar spine MRI reports.
- 125 reports (104 human, 21 AI) were blindly evaluated by 5 medical professionals.
- Radiologist reports scored higher in clinical relevance, clarity, completeness, and accuracy.
- No clinically false statements were found in AI-generated reports.
- Identification accuracy varied among evaluators, with radiologists being most accurate.
- AI-generated reports were sometimes indistinguishable from human ones, especially for non-specialists.
- LLMs may assist radiologists in structured reporting and improve workflow efficiency.
Pentagon officials send Anthropic best and final offer for military use of AI
2 months ago
- Pentagon officials sent Anthropic a best and final offer for military use of its AI technology ahead of a government-imposed deadline.
- Defense Secretary Pete Hegseth set a Friday deadline for Anthropic to grant full lawful use of its AI or lose U.S. military business.
- Anthropic could face being labeled a supply chain risk if it does not comply with the Pentagon's demands.
- The Pentagon is considering invoking the Defense Production Act to gain full control of Anthropic's AI for military operations.
- Anthropic was awarded a $200 million contract in July to develop AI capabilities for U.S. national security.
- Anthropic has requested guardrails to prevent its AI model, Claude, from mass surveillance of Americans, which the Pentagon says is already illegal.
- Anthropic's CEO, Dario Amodei, wants to ensure Claude is not used for final targeting decisions in military operations without human involvement due to potential AI hallucinations and errors.
Human brain cells on a chip learned to play Doom in a week
2 months ago
- A clump of human brain cells can play Doom, marking progress toward biological computers.
- Cortical Labs developed neuron-powered chips that played Pong in 2021 and now Doom using Python.
- The Doom-playing chip learned faster than silicon-based systems but performed below human levels.
- Experts highlight the potential for real-world applications like controlling robotic arms.
- The advancement shows biological systems can handle complexity and real-time decision-making.
Reporting checklist for foundation and large language models in medical research (REFINE): an international consensus guideline - PubMed
2 months ago
- Development of the REFINE checklist for transparent and reproducible reporting of foundation and large language models in medical research.
- Modified Delphi process involving 57 contributors from 17 countries to establish reporting standards.
- Resulted in a 44-item, six-section framework with standardized terminology and detailed reporting instructions.
- Online platform provided for practical implementation of the REFINE checklist.
- Aims to support reliable evidence synthesis in medical and imaging-focused AI studies.
The role of artificial intelligence in advancing population-based cancer registration - PubMed
2 months ago
- Cancer is the second leading cause of death globally, with increasing burden and disparities.
- Population-based cancer registries collect data crucial for cancer prevention and control strategies.
- Artificial intelligence (AI) can reduce the labor-intensive nature of registry operations.
- AI enables processing large datasets and extracting complex or unstructured data patterns.
- AI supports data quality control and enhances cancer registration data abstraction.
- AI methodologies are increasingly integrated into the analysis and dissemination of registry data.
- Challenges include network and computational constraints, uneven resource allocation, and AI biases.
- A forward-looking AI-enhanced framework is proposed to optimize cancer registration efficiency.
- AI can enhance the use of registry data for cancer control and research.
Exosome-enabled bone defect repair: mechanistic foundations, bioengineered delivery, and artificial intelligence-driven translation - PubMed
2 months ago
- Exosomes are promising cell-free therapeutic candidates for bone defect repair due to their biocompatibility.
- Mechanisms of bone regeneration include osteogenic and angiogenic activation, osteoblast-osteoclast coupling modulation, immune regulation, and extracellular matrix remodeling.
- Engineering strategies improve exosome targeting, stability, and potency, while delivery platforms offer controlled release at defect sites.
- Artificial intelligence (AI) accelerates translation through exosome characterization, biomaterial design, and therapeutic response prediction.
- Challenges include heterogeneity in isolation, low yield, lack of standardization, and insufficient validation in disease-relevant models.
- Future directions focus on standardized manufacturing, mechanism-informed engineering, responsive delivery systems, and AI-driven design-control pipelines.
The Lost Aura of the Physician in the Age of Artificial Intelligence - PubMed
2 months ago
- The article discusses the evolving role of physicians with the rise of artificial intelligence in medicine.
- It highlights how AI is impacting the traditional aura and authority of physicians.
- The piece reflects on the challenges and changes in the medical profession due to AI integration.
This job has become the ultimate case study why AI won't replace human workers
2 months ago
- AI is transforming radiology by enhancing efficiency without replacing human radiologists.
- Radiology is ideal for AI due to abundant digital data and the need for quick analysis.
- AI helps prioritize scans, improve image quality, and assist in report summaries.
- Human radiologists remain essential for diagnoses, patient examinations, and report writing.
- Radiology jobs are projected to grow 5% from 2024 to 2034, faster than the average occupation.
- AI tools must undergo FDA approval, which can take around eight years.
- Over 1,000 FDA-approved AI-enabled medical devices are for radiology.
- AI's role in radiology was initially feared but is now seen as a supportive tool.
- Risks include bias in AI diagnoses and potential overreliance on technology.
- Collaboration between AI and human experts is crucial for accurate and improved outcomes.
BMW Group to deploy humanoid robots in production in Germany for the first time
2 months ago
- BMW Group introduces humanoid robots in production at Leipzig plant, marking the first deployment in Europe.
- Physical AI combines digital AI with real machines, enhancing production processes and competitiveness.
- A pilot project in Spartanburg, USA, successfully tested humanoid robots, moving over 90,000 components in 1,250 operating hours.
- The BMW Group establishes a 'Center of Competence for Physical AI in Production' to consolidate expertise.
- Humanoid robots are seen as complementary to existing automation, especially in monotonous or safety-critical tasks.
- Collaboration with Hexagon Robotics for the European pilot project, focusing on multifunctional applications in battery assembly and component manufacturing.
- Integration into production IT infrastructure and employee acceptance were key factors in the successful Spartanburg pilot.
- Future plans include evaluating additional use cases for humanoid robots in production.

first prev10next

About|Login

#artificial intelligence

Development and Validation of a Computational Histology Artificial Intelligence-Powered Predictive Biomarker for Selection of Chemotherapy in Advanced Pancreatic Cancer - PubMed

Enhancing therapeutic outcomes with artificial intelligence for HR-positive, HER2-negative metastatic breast cancer - PubMed

Artificial intelligence agents in healthcare research: A scoping review - PubMed

Study: Self-generated Agent Skills are useless

How AI is affecting productivity and jobs in Europe

Large Language Model Reasoning Failures

Artificial intelligence and digital health in heart failure: advances in diagnosis, monitoring, phenotyping, and digital biomarkers - PubMed

Transforming nephrology through artificial intelligence: a state-of-the-art roadmap for clinical integration - PubMed

Research priorities for data science and artificial intelligence in global health: an international consensus exercise - PubMed

Artificial intelligence-generated marathon training programs: reliable tools in exercise prescription for athletic performance? - PubMed

Artificial intelligence-generated synthetic data for cancer research and clinical trials - PubMed

Can AI write reports like a radiologist? A blinded evaluation of large language model-generated lumbar spine MRI reports - PubMed

Pentagon officials send Anthropic best and final offer for military use of AI

Human brain cells on a chip learned to play Doom in a week

Reporting checklist for foundation and large language models in medical research (REFINE): an international consensus guideline - PubMed

The role of artificial intelligence in advancing population-based cancer registration - PubMed

Exosome-enabled bone defect repair: mechanistic foundations, bioengineered delivery, and artificial intelligence-driven translation - PubMed

The Lost Aura of the Physician in the Age of Artificial Intelligence - PubMed

This job has become the ultimate case study why AI won't replace human workers

BMW Group to deploy humanoid robots in production in Germany for the first time