Learning center

SUV preparation: units, region review and export scope

Review quantitative PET inputs, decay references, partial results and the difference between a results bundle and a viewer volume.

Source-guided preparation and export workflow. Medical Image Processing SUV Calculation, medical_image/suv_calculation, requires quantitative PET evidence before a number is meaningful. This guide explains the current input, region and export contracts using retained regression evidence. It adds no new scan, independently matched SUV result or GUI recording.

1. Establish what the pixels represent

Keep the original acquisition, its metadata and the loader’s source identity together. Open SUV Calculation in Nuclear Medicine after loading the intended PET input. A PET-looking image, a filename or a manually edited modality label does not supply quantitative calibration. The current source refuses contradictory encoded modality, unsupported or conflicting units, and pixels with an unspecified value domain.

EvidenceWhat to preserve and review
Pixel units and domainActivity concentration must have a supported quantitative unit and modality/real-world value domain. Stored counts are insufficient. The app distinguishes BQML from already encoded GML SUV data.
Normalization basisKeep the SUV Type with the image. Body weight, body surface area and lean body mass are different normalizations. The current SUL (Janma) route uses the Janmahasatian basis; do not relabel a James-based value as the same method.
Administered activityFor conversion from activity concentration, retain the supplied activity unit, relevant administration time and residual-activity evidence. The core distinguishes Bq from MBq; it does not infer a missing unit from the number’s size.
Frame and geometryRetain axis order, frame count, temporal meaning, timing units and physical voxel geometry. Select the intended frame explicitly; a volume axis is not automatically time.

The DICOM PET Series definition distinguishes SUV Type from Units and defines a body-weight default when SUV Type is absent for GML data. The current app follows that default. For a source already in the requested SUV basis, it preserves the encoded values; conversion to another basis uses the normalization ratio. It does not apply an injected-dose normalization again to matching-basis GML input.

For PET-BIDS, keep the associated JSON with its image. The PET-BIDS specification defines explicit activity units and timing relative to TimeZero. The current app accepts supported Bq/mL, kBq/mL and MBq/mL scaling only through its loader-verified PET-BIDS contract. Editing a unit label alone does not establish that contract or full BIDS conformance.

2. Keep activity and pixels on the same time reference

For a BQML conversion, the app scales by the selected body normalization and the administered activity at the image’s decay reference. For body weight, that means weight in grams divided by activity in Bq, applied to concentration in Bq/mL. Keep the output’s declared basis and units with the result. This describes the source calculation, not an independent accuracy check.

ReferenceCurrent source interpretation
ADMINActivity is referenced to administration; no elapsed decay interval is applied to the supplied administration-referenced activity.
STARTUse the explicit interval from administration to acquisition start. A nonzero interval also needs the radionuclide half-life.
NONEUse an explicit effective time for each frame. DICOM Frame Reference Time is relative to Series Reference Time; a scan-start interval cannot silently substitute for that different origin.
Enhanced-PET date-timeThe app’s separate DATETIME branch uses the declared correction date-time relative to administration. It is not an extra value to invent in the classic DICOM Decay Correction field.

DICOM’s Decay Correction definition supplies the ADMIN, START and NONE context. Missing or contradictory timing needs review of the acquisition evidence. Changing a timestamp to make a calculation proceed would change the scientific meaning.

3. Review sidecars, regions and selected analyses

  1. Use LOAD QUANTITATIVE SIDECAR only for evidence belonging to the loaded acquisition. EXPORT SIDECAR TEMPLATE creates a source-bound JSON template whose measurements are empty and whose status is unvalidated. Populate it from verified records and review it; the template is not a calibration certificate.
  2. Set Scope to Selected frame or All frames deliberately. The selected-frame control starts at index 0. The source bounds all-frame metric work to 512 frames; lazy volume/export routes have their own limits. Those limits do not establish available hosted capacity.
  3. For regional results, use LOAD ROI / LABEL MASK and PREPARE / REVIEW ROIS. Choose the actual role: lesion, reference, background, analysis or PVC labels. Review the mask on the matching physical grid. An array with the same dimensions can still have the wrong position or orientation.
  4. Choose the needed analyses and inspect their prerequisites. Threshold outputs are labeled candidates. A default threshold or a protocol label is not an independently established lesion boundary, clinical recommendation or QIBA/PERCIST conformance result.
  5. Read the run and route statuses together. An available volume can coexist with an unavailable region-dependent result. Preserve the reason instead of reporting an unavailable value as zero.

4. Read a partial result honestly

A retained synthetic regression selects suv_bw and suv_peak without supplying the peak route’s required region. Its recorded expectations are a partial run, successful SUVbw and unavailable SULpeak (the suv_peak route). This is a narrow control-flow example from the saved test receipt, not a patient result or a new numerical comparison.

The retained selected suite reports 13 passed and 2 skipped. The full FXS2 launcher test was opt-in and skipped; the official QIBA DRO launcher test was skipped because the DRO was not downloaded for that run. The SUV app, five support files and the test file still match that receipt’s source hashes. The current launcher differs, so these historical results do not establish a new current-launcher workflow acceptance. Twenty route labels in current source are not twenty validated methods.

5. Save the right object and its context

Export bundle can perform analysis. If no volume result exists, the current route creates an SUVbw volume for export using the selected scope. Choose it intentionally after reviewing input and route prerequisites. The bundle’s member counts depend on the results available at export time; this guide does not promise a fixed number of files.

ExportContents and boundaries
Results bundleAvailable volume NPY files, verified ROI/label arrays, complete available result tables as CSV, generated figures as PNG, and a manifest with member sizes and SHA-256 hashes. The manifest retains source contract, units, geometry, quantitative QC, preparation and run context. A route that was unavailable has no successful numerical result to export.
Viewer volumeA chosen .npy file and paired .npy.json metadata. Keep both: axis order, timing, geometry, units and the array hash are in the sidecar. Existing destinations are refused. ROI overlays are not embedded in this array export; their role names are recorded separately.
Numerical precisionBoth volume exporters write float32 arrays. CSV exports use the complete stored table rows rather than a cropped viewer image. An exported float32 value need not retain every digit of an internal calculation.
Export summaryThe Export bundle route adds its summary table after writing that bundle. Do not expect the newly added export-summary row inside the bundle it just described. Inspect the manifest’s actual members and retain the later run display separately when needed.

The source retains computed_not_independently_validated and publishable_as_validated = false in export metadata. A completed export preserves a result; it does not promote its scientific status. This guide checks the export contract in source and a historical paired-export regression. It does not claim a newly produced or independently compared SUV bundle, a complete saved session, or a verified re-open of the original acquisition.

Worked examples describe particular inputs, methods and recorded results. Check current availability and input requirements before planning an evaluation.

← Back to the learning center