Skip to content

QA test plan

This plan lists the behaviours of SIV's visualization features as numbered test cases, one behaviour per case, each with its preconditions, steps and expected result. It is written for the SQA team to verify a release, but anyone can use it to check that a feature works as documented.

The expected results describe what the current code does. Where the code and earlier documentation disagree, or where a behaviour looks like a defect, the case is tagged ⚠ see Known issues and is listed on Known issues to confirm. Record what you observe; do not fail a case for a behaviour that page already describes, unless it changed.

Test areas

Page IDs Covers
Viewing VIEW-001 … Toolbar, plot types, navigation, Channels & Histogram, stacks, 3D, slice editor, profiles, export
Regions REG-001 … Drawing and editing tools, selection, undo/redo, context-menu region actions, shortcuts
Region Editor and classes RED-001 … Region Editor table and toolbar, classes and presets, GeoJSON/mask import and export, 3D classes
AI segmentation SEG-001 … SAM box and point prompts, model picker, Cellpose, progress toasts, model caching
Spatial omics SPA-001 … Spatial omics 2D/3D, panel, colouring, selection, distribution charts, transcripts, fallbacks
Known issues to confirm KI-01 … Suspected defects and documentation drift, with source references

How to use the plan

Test IDs

  • Every case has a stable ID: an area prefix and a three-digit number (REG-014).
  • IDs are never renumbered. When a behaviour is removed, the case is marked Retired in x.y.z and keeps its number; new cases take the next free number at the end of the page, even if they belong to an earlier section. Results from different releases can therefore be compared ID by ID.
  • Known issues use KI-nn and follow the same rule.

Priorities

Priority Meaning When to run
P1 Core path: if it fails, the feature is unusable. Together the P1 cases form the smoke test. Every release candidate, every browser
P2 Important behaviour, defaults, limits and error messages users rely on. Every minor release, main browser
P3 Edge cases, cosmetic details, rarely used options. Major releases, or when the area changed

Recording results

Use one of these outcomes per case and per browser:

Outcome Use when
Pass The observed behaviour matches the expected result.
Fail It does not. Note what you saw, attach a screenshot, and open an issue.
Blocked The case cannot be run (missing data, a preceding failure, browser lacks WebGPU). Say why.
N/A The case does not apply to this environment (for example a WebGPU-only case in the fallback run).

Copy this template for a test run:

## SIV x.y.z — test run <date> — <tester>

Environment: <browser + version> · <OS> · WebGPU: yes/no · <screen size> · <demo or JIT build>

| ID       | Outcome | Notes / issue link |
| -------- | ------- | ------------------ |
| VIEW-001 | Pass    |                    |
| VIEW-002 | Fail    | #123 — …           |

Test environment

Browsers

Browser WebGPU Role
Chrome (latest) — desktop Yes Primary. Run every case.
Edge (latest) — desktop Yes Run P1 and P2.
Firefox (latest) From version 141 on supported systems Run P1; repeat the WebGPU-dependent cases if WebGPU is available.
Safari (latest) From version 26 Run P1.
Any browser, WebGPU off No Run the fallback cases (SPA- fallback section, VIEW- 3D section).

Running without WebGPU

Several features (Surface, Volume, Isosurface, Spatial omics 3D, napari rendering) need WebGPU, and Spatial omics 2D falls back to a WebGL drawing without it. To test the fallback paths:

  • use a browser or system without WebGPU (the viewer asks for a WebGPU adapter once per page, with a 3 s timeout); or
  • in Chrome/Edge, open chrome://flags (edge://flags), set Unsafe WebGPU Support / WebGPU-related flags to Disabled, or start the browser with --disable-features=WebGPU, and restart; or
  • serve the page over plain http:// from a non-localhost host (WebGPU needs a secure context).

Check navigator.gpu in the developer console: undefined means no WebGPU. Backend fallbacks are reported in the console only (for example [visualizer] spatial view drawn with WebGL (no WebGPU): …), so keep the console open during fallback runs.

Screen sizes

Size Purpose
1920 × 1080 Reference layout.
1440 × 900 Typical laptop; toolbar wraps to two rows.
1280 × 720 or less Dialog widths switch to 90vw below 960 px; toolbar wraps further.
High-DPI (Retina) Brush, eraser and wand sizes (readback in device pixels), screenshot crispness.

Test data

Most cases use the live demo, https://thejacksonlaboratory.github.io/sci-image-visualizer/, which shows a gallery of samples on the left and the viewer on the right. The whole-slide (WSI) entries and the spatial-omics folder are served by the JAX development tile server and need network access to it.

Gallery entry Type Exercises
img01.png Small 8-bit image (muscle section, grey) Region tools, wand, brush, region actions, Heatmap, intensity profiles
img03.png, 052_img.png, 024_img.png Small fluorescence / cell images Wand on bright objects, SAM point prompts
img04.png RGB fluorescence (blue nuclei, green cells) RGB channels, Channels & Histogram, SAM box prompt, Cellpose
002_img.png RGB, many cells Cellpose, many-region performance
031_img.png, 034_img.png RGB, green structures Wand, Contour availability on RGB (must not be offered)
BF-FL_composite.png Large PNG composite (~12 MB) Serverless large image, zoom, smoothing
TIFF B16-69_O7_CFP-1_hyper.tif TIFF decoded in the browser Client-side TIFF, multichannel or hyperstack handling
TIFF GFPA_IBA1_…_zstack_Series002.tif Multichannel z-stack TIFF Slice slider, ← / → keys, Volume/Isosurface, channel visibility
micro-ct folder (DCM slices) Grayscale DICOM; right-click = z-stack Single DICOM slice, Contour, colormaps, z-stack, Volume, Isosurface, slice editor, Heatmap stack mode
WSI CMU-1 · 1.5 Gpx (CC0) Gigapixel H&E, tiled, with pixel size Tiled loading, minimap, scale bar, µm² areas, patho-sam, export composite
WSI BC18 · 22 Gpx (NDPI) Very large NDPI Tile loading at scale, region sources
WSI Sirius Red · 0.5 Gpx (NDPI) Brightfield stain Wand on stained tissue
WSI 2-channel fluorescence · 0.35 Gpx Multichannel, tiled, >8-bit Channel table, LUT row, native histogram, 16-bit TIFF export
WSI 2-channel · multichannel:false (repro) Regression sample Multichannel flag handling
WSI Project002 · 2-channel BF · 0.28 Gpx (z=13) Tiled z-stack Slice slider on a tiled stack, preserveViewport
WSI Project002 · 2ch x 27z stack · 7.5 Gpx Large tiled z-stack Stack loading, Cancel, Volume
spatial-omics → Adult mouse brain Visium H&E 2,688 spots, 18,078 genes, H&E image Spatial omics 2D, colour by group/gene/measurement, Select from ROIs, all distribution charts
spatial-omics → SYNTHETIC demo-brain ~2,000 spots, synthetic H&E, fabricated values Spatial omics 2D (small, fast)
spatial-omics → Mouse embryo seqFISH 19,416 cells, no image, 6 embeddings Dataset without image (pixel modes withheld), Embedding chart, lasso
spatial-omics → Allen Brain Cell Atlas MERFISH 3D, 3.7 M cells (and 1-in-10 subset), CCF volume Spatial omics 3D, volume, class tree, class-volume selection, large data

What the demo cannot test

The demo server has no transcript (Xenium) dataset, no gene-heatmap All genes data, no cell segmentations, no server region import/save and no annotation-class persistence (PREFERENCES_PORT). Run those cases in JAX Image Tools (JIT), which provides the server side. Cases that need JIT say so in their preconditions.

Smoke test

The smoke test is every P1 case. In order:

  1. Viewing: all P1 VIEW- cases (open each sample type, plot types, navigation, channels, stack, 3D).
  2. Regions: all P1 REG- cases (draw with each tool, select, move, delete, undo/redo).
  3. Region Editor: all P1 RED- cases (table, classes, GeoJSON export and re-import).
  4. Segmentation: all P1 SEG- cases (SAM box, SAM point, Cellpose once each).
  5. Spatial omics: all P1 SPA- cases (open a dataset, colour it, select from a region, show a chart).

Run the smoke test in the primary browser and, for the WebGPU-dependent P1 cases, once with WebGPU off.