---
title: "2026-08-30 NanoVirDx research loop figure"
description: "Interpreted the NanoVirDx pseudo-schema and added a circular figure for its core modules and local optimization loops."
image: "https://docs.virdx.dev/img/virdx-social-card.png"
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.virdx.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# 2026-08-30 NanoVirDx research loop figure

## Work Done

- Read `packages/nanovirdx/src/nanovirdx/main.py` as a conceptual schema spanning histology embedding, MRI simulation, and inversion.
- Identified the central path `WSI -> DomainRepresentation -> synthetic MRI -> recovered tissue/clinical state`, with reconstruction or task loss closing the cycle.
- Distinguished the four parameter families and their local optimization scopes: embedding, tissue calibration, acquisition sequence fitting, and inverse training.
- Added an editable SVG at `packages/nanovirdx/docs/research-loop.svg` and interpretation notes at `packages/nanovirdx/docs/research-loop.md` in the mono workspace.
- Validated the SVG as well-formed XML.

## Lessons Learned: Pitfalls

- The source calls the sequence loop “optimization,” but its sketched loss only matches a real MRI acquired with fixed settings. This is acquisition fitting, not yet task- or information-optimal sequence design.
- The comments state an optimization order (`EmbeddingParams -> TissueParams -> SequenceParams -> InverseParams`), but this should not be presented as a causal dependency graph.
- The pseudo-code contains unresolved names (`_params_tissue`, `_mri_acq_params`) and an unimported `Any`; these were treated as sketch-level inconsistencies rather than encoded into the figure.

## Lessons Learned: Improvements

- A durable NanoVirDx design note should define the boundary between `DomainRepresentation` and `TissueParams`, the actual loss/registration spaces, and whether inversion targets WSI reconstruction or clinical endpoints.
- Sequence design needs an explicit downstream utility or information objective plus physical acquisition constraints before it can be shown as true sequence optimization.
- Paired WSI/MRI evidence, WSI-only synthetic training, and real-MRI supervised training should stay visually separate because they support different loops and carry different assumptions.

Source: https://docs.virdx.dev/knowledge/inbox/2026-08-30-diffsim-nanovirdx-research-loop-figure/index.mdx