30-Second Takeaway
- Multisociety Delphi defines when radiography is appropriate for 12 low-yield anatomic sites.
- Explainable AI in cancer imaging remains fragmented with limited quantitative validation and deployment.
Week ending May 23, 2026
Grand Rounds: Recent evidence on imaging equity, radiography appropriateness, xAI, trial labeling, and AI-driven mammography workflow
Pediatric ED imaging disparities persist despite pediatric capability
This retrospective cohort of 857,034 pediatric ED visits found public insurance and non-Hispanic Black or Hispanic children were less likely to receive imaging across chest, head, and abdominal indications. Measures of hospital pediatric capability (pediatric emergency care coordinator, NPRP readiness, bed capability) influenced imaging frequency but did not change disparities by insurance or race/ethnicity. Findings were consistent in sensitivity analyses limited to discharged patients. Implication: institutional pediatric resources alone do not eliminate imaging inequities.
Multispecialty Delphi consensus on when radiography is low-yield
A 34-member multidisciplinary panel produced 58 consensus statements covering radiography for 12 low-yield anatomic sites. Statements were derived from a structured literature review and three Delphi rounds and received endorsement from six specialty societies. This provides explicit, society-endorsed guidance to reduce unnecessary radiographs for specified sites.
Explainable AI in cancer imaging is heterogeneous and under-validated
Scoping review of 371 studies found xAI dominated by post hoc visualization and feature relevance methods (82.2% post hoc). Most work used deep learning (70.1%) with few studies providing code (17.5%) or integrating into decision support (12.1%). Quantitative validation was rare and clinical deployment was uncommon, limiting immediate clinical trust and use.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.