Figure 1a-h | Agent-orchestrated tools for reconstructing and reviewing anatomical atlases.

a, Curated categorical reference sections, regional identities, physical scale and section ordering enter the workflow. b, Image-processing tools extract contours from the curated maps. c, Extracted regions receive human-supplied identities from the curated registry; the agent does not infer anatomical identities. d, The parallel-section workflow applies bounded in-plane rigid alignment using shared regional geometry. e, Categorical reconstruction interpolates between aligned anchors using shared-map warping with signed-distance interpolation, with a distance-based fallback when the geometric checks require it. A denotes a retained anchor plane in the schematic. f, The deterministic tools export a labelled volume, surfaces, an interactive viewer and geometric QA/provenance records. A user-configured agent API calls registered tools and may propose bounded corrections; dependency checks and measured geometric criteria govern their application. Bidirectional links indicate human inspection and correction of boundaries, identities, spatial alignment and reconstructed outputs. Geometric acceptance does not establish anatomical validity.

g, Coronal, horizontal, sagittal and oblique reference planes can be supplied as parallel series with one reviewed series-to-world 4 x 4 transform per entire series and a shared, human-curated identity registry. Each transform applies to local XY coordinates and the individual section Z coordinates within that series. h, These supplied placements define a common coordinate frame for conservative multi-plane integration. Agreeing labels retain source evidence, whereas conflicting labels remain unresolved (65535). This integration does not perform autonomous cross-plane anatomical registration; spatial consistency and final anatomical acceptance require human review.

All drawings in panels a-h are conceptual vector schematics, not empirical anatomical reconstructions or measured plots. Existing empirical panels i-q are outside this replacement and remain unchanged.
