Glossary · Neurovia

C

Clinical Decision Support (CDS): System capabilities that provide recommendations, alerts, or structured summaries during care to support faster, more consistent decisions—not to replace physician sign-off.

DICOM: International standard format and communication protocol for medical imaging. CT, MRI, and other studies typically move between hospital systems as DICOM.

EHR: Electronic Health Record—the institution’s longitudinal clinical record system (orders, notes, results, and care documentation). In the U.S., AI products typically integrate with EHR and PACS rather than replacing them.

FHIR: A modern healthcare data exchange standard that makes it easier for systems to share patient events, reports, and workflow information.

HL7: Family of healthcare messaging standards, commonly used for exam orders, report status, and other event messages.

Large Language Model (LLM): AI that understands and generates human language. In healthcare it is often used for report drafts and tumor board summaries—always with human review.

Tumor board: In U.S. oncology practice, a multidisciplinary case conference (sometimes called an MDT elsewhere) where specialists jointly review complex cases. AI can shorten pre-meeting prep with evidence packs and summaries.
Machine Learning (ML): Techniques that let computers learn patterns from data and improve performance. Medical image recognition and risk scoring often rely on ML models.
Multimodal fusion: Aligning and jointly analyzing information from imaging, pathology, omics, and other sources to build a more complete evidence chain.

Natural Language Processing (NLP): Technology for understanding, extracting, and generating human language—often used to pull key facts from report text.

PACS: Picture Archiving and Communication System—stores and retrieves medical images. Imaging AI ideally embeds in existing PACS workflows.
On-premises deployment: Software and models run in the organization's own data center or private network; data stays in-domain by default—suited to high-compliance settings.

Risk stratification: Sorting cases into actionable tiers (e.g. surveillance / specialty review / tumor board prep) by risk level, with rationale where possible.

Explainable AI (XAI): Methods and interfaces that make model outputs understandable—feature contributions, saliency cues, confidence intervals—to support QA and trust.