LAUDOS.AI RESEARCH · INTERACTIVE PAPER · DRAFT, MAY 2026

Beyond Templates

Why the radiology report is not a template-selection problem, and what a reporting system that truly scales looks like. By Natan Paraíso Ribeiro, MD; Petrus Paraíso Ribeiro, MD; Stephanie Alba Herrera, MD; and Francisco Akira, MD.

Abstract: every signed report is rendered from a clinical state with multiple dimensions: indication, technique, comparison, findings, measurements, impression and the signer's style. Template libraries try to pre-write outputs for this space, but the space grows by multiplication of dimensions while catalogs grow one template at a time. The work formalizes this asymmetry and describes the alternative implemented at Laudos.AI: a compositional engine that stores reusable semantic components and renders the report at the moment of signature, with the template demoted to the role that fits it, that of a presentation surface. The paper is interactive: the arguments can be explored, slid and composed on the page itself.

The interactive version loads on this page with JavaScript enabled. Published by Laudos.AI Research, São Paulo, Brazil. See also: RadCommons · LaiBench · Methodology · LaudAI.