Artificial intelligence

Where the papers sit

89 papers from the last 6 months study artificial intelligence directly · updated from PubMed 10 Oct 2026

• Medical AI Education and Governance

  Medical AI is moving beyond model validation toward lifecycle surveillance, regulation, workforce education, and patient-centered communication. Post-deployment monitoring, adaptive governance, and accessible training recur across clinical and academic settings. Proof-of-concept tools for heart failure and aortic measurement show continuing clinical development, while generated summaries expose speech and speaker-related errors. Trust, nursing judgment, and community participation remain open requirements.

  17 papers · 19.1%

• AI in Diagnostic Imaging

  Diagnostic AI is moving from retrospective accuracy claims toward clinical utility in imaging, pathology, and risk stratification. Lung nodule reading, prostate MRI, thyroid ultrasound, pancreatic risk, and autoimmune hepatitis recur across the work. Sensitivity, specificity, predictive values, reader time, and radiation toxicity prediction anchor evaluation. Generalization across images and prospective validation remain unresolved.

  16 papers · 18%

• Clinical AI Validation and Triage

  Emergency triage models such as START-AI are moving toward admission-risk prediction and specialty assignment in routine care. Validation spans self-triage, readmission prediction, diabetic retinopathy screening, echocardiography, and pathology. Cost, sex bias, safety monitoring, and clinician oversight recur as barriers to deployment. The evidence supports practical augmentation, but not autonomous decision-making.

  15 papers · 16.9%

• Trust and Adoption in Healthcare

  Safe adoption of AI is being framed around patient trust, clinician judgment, and equitable access rather than accuracy alone. Shadow AI, ambient ICU scribes, and ChatGPT-based care raise concerns about privacy, accountability, and health disparities. Cost-utility work suggests value when tools improve outcomes or reduce workload. Literacy, human factors, and implementation support remain prerequisites for sustained use.

  14 papers · 15.7%

• Healthcare Workforce AI Readiness

  Healthcare AI adoption is being shaped by workforce training, readiness, trust, and local implementation conditions. Mixed-methods studies examine primary care, obstetrics, radiology, pharmacy, nursing, and older-adult care, while Zambia provides a point-of-care ultrasound deployment example. Education in retrieval-augmented generation and AI agents is expanding alongside practical tools such as fracture detection and pill recognition. Sustained use depends on competence, workflow fit, and confidence rather than access alone.

  12 papers · 13.5%

• Clinical AI Evaluation and Implementation

  Clinical AI work is emphasizing safe implementation, user acceptance, and decision support across mental health and broader care settings. Psychotherapy agents are being stress-tested in high-risk scenarios, while consumers and clinicians report differing expectations of benefit and harm. Quality-assurance frameworks and shared decision-making studies push toward accountable human oversight. Evidence remains exploratory, and real-world effectiveness is less established than perceived usefulness.

  9 papers · 10.1%

• Large Language Models in Care

  Large language models are being tested for bounded clinical work, including prescription review, wound assessment, psychotherapy support, and resident microlearning. Multimodal systems and locally deployed models aim to improve capability while limiting privacy and surveillance risks. Human supervision remains central in therapy and Socratic medical education. Evidence is still early, with feasibility studies and cost-effectiveness modeling preceding outcome trials.

  6 papers · 6.7%