Color Doppler Looks Wrong in PACS: Check the Pixel Encoding

Why can the same ultrasound file look different in two viewers?

A useful starting observation is specific: the same saved frame has one appearance in the scanner archive and a different appearance in the receiving viewer. Keep the frame number, visible color bar, object identity and display versions together. Comparing two different frames, a live image against a saved still, or a screenshot against a DICOM object can produce a misleading comparison before any decoder fault is involved.

The exported image may already contain rendered anatomy, annotations and a color scale. DICOM then describes how its stored pixel samples should be interpreted. Those samples may be direct RGB components, luminance and chrominance components, or palette indices. They are not necessarily raw ultrasound acquisition-channel data or separately recoverable numerical velocity measurements. A display discrepancy therefore cannot establish that a clinical measurement changed. ^{1 2}

The immediate decision is who should investigate next. Avoid selecting a replacement board from the hue of an exported image. First establish whether the discrepancy follows one file, one viewer build, one receiving route, or the scanner itself. This is an evidence sequence, not a claim that most faults have one cause: the cited support cases do not provide a denominator for fault frequency.

Related delivery failures need different checks. If the viewer finds no study at all, use the USB DICOMDIR and viewer-loading guide. If the concern is a distance measurement after export, the region-calibration guide addresses that separate requirement. Neither directory recovery nor calibration metadata substitutes for checking pixel color encoding.

Which attributes belong in the first evidence record?

Photometric Interpretation (0028,0004) declares the intended pixel interpretation. Samples per Pixel (0028,0002), Planar Configuration (0028,0006) and Bits Allocated (0028,0100) help describe the representation. Record their actual values, including an absent conditional field, rather than filling in what a troubleshooting worksheet expects. The table below is a recognition aid for conventional US objects, not a compatibility list for a particular scanner or viewer. ^{1 2}

For three-component data, planar value 0 describes color-by-pixel ordering, while 1 describes color-by-plane ordering. Do not use those labels as a manual description of JPEG compressed bytes: the codec controls the compressed representation. For single-sample monochrome or palette images, Planar Configuration is not applicable and must not be inserted as a supposed repair. ²

Read the ultrasound specialization as a whole. Its transfer-syntax rule associates uncompressed color with RGB, JPEG lossy color with YBR_FULL_422, and RLE with RGB or YBR_FULL. The planar table also contains YBR_FULL rows for other encodings, while the general Image Pixel text contains a restrictive US cross-reference. These passages are not evidence that any arbitrary combination is a supported export. Where declarations appear inconsistent, preserve them and request a conformance review against the exact object, syntax and supplier statement; do not resolve the discrepancy by editing a tag. ^{1 2}

Ultrasound Color Data Present (0028,0014) is optional, Type 3. If supplied, 0 indicates no ultrasound color data and 1 indicates its presence. Absence means unknown from this field, not no color. The standard also permits an RGB image with equal R, G and B samples throughout. This explains a valid grayscale RGB presentation; it does not explain why a particular expected Doppler frame is missing its color. ¹

What can go wrong when YBR is treated as RGB?

YBR separates brightness from color-difference components; RGB uses red, green and blue components. Sending a decoded YBR array straight into a display API expecting RGB can shift apparent colors. The observed hue is not a reliable signature for one defect, however. A mislabeled dataset, an unexpected decoder output space, an inappropriate second conversion or display settings can require different corrections. Record the decoding pipeline before changing it. ^{2 3}

The 4:2:2 suffix concerns chroma subsampling. For native 8-bit YBR_FULL_422, two adjacent luminance samples share chroma samples, with the documented order Y1, Y2, Cb, Cr. That is four sample bytes for two pixels rather than six for native 8-bit RGB. This native layout is not the physical layout of an encapsulated JPEG fragment. The standard notes that actual JPEG bitstream subsampling may differ from the photometric label; the decoder must inspect the compressed representation. ²

A practical software handoff therefore names the library, its version, decoder plugin, input transfer syntax and output-array color space. It also records whether the application applies a color transform after decoding. Without that sequence, a request to add YBR-to-RGB conversion could apply the transform twice. Investigate with authorized test copies rather than rewriting clinical archive objects.

Historical documentation needs a version boundary. The 2022 pydicom discussion describes unexpected colors and a possibly mismatched RGB declaration. Current pydicom 3.0.2 documentation says its pixels.pixel_array function converts YBR_FULL and YBR_FULL_422 to RGB by default; raw=True changes the processing requested. An old statement that pydicom never converts automatically is therefore unsuitable as present-day advice. ^{3 6 4 5}

What must be checked for PALETTE COLOR?

A palette image stores one index per pixel. To obtain display colors, the receiver needs the associated lookup information. Record Red, Green and Blue Palette Color Lookup Table Descriptors (0028,1101), (0028,1102), (0028,1103), together with their Data elements (0028,1201), (0028,1202), (0028,1203). Where segmented palette data is used, retain those elements too and have the receiving implementation check support. Descriptors alone do not contain the full palette. ²

The descriptor identifies the entry count, first mapped input and bits per entry. An entry-count value of zero represents 65,536 entries, rather than an empty table. Capture the descriptor values and data lengths for specialist review, without pasting patient-identifying headers into a public support forum. Do not replace a missing table with a visually plausible red/blue gradient: that would create a new mapping rather than recover evidence. ²

A viewer opening the file elsewhere makes a receiver-specific issue plausible, but cannot prove conformance or reveal the original failure mechanism. A tolerant viewer might compensate for an inconsistency that another rejects. Likewise, the presence of palette tags does not prove their lengths and mappings are valid. The receiving vendor should examine the complete object and the declared capability for that SOP class, syntax and palette representation.

Do not infer clinical flow direction from a palette index, monitor hue or the cover illustration. An image may carry a rendered color scale, but the relationship between displayed color and any quantitative ultrasound information needs its own source evidence. A color-encoding investigation must leave clinical interpretation with the responsible clinical team.

How can you compare objects without overwriting the evidence?

1. Preserve a bounded comparison set. Keep the original export and a separately retrieved archive copy under facility access controls. Record where and when each came from, how it was retrieved and the selected frame. Some retrieval paths transcode on demand, so distinguish a downloaded representation from the archive’s stored representation. Screenshots record an appearance, not pixel-data identity.

• Record SOP Class UID (0008,0016), SOP Instance UID (0008,0018), frame number and Number of Frames (0028,0008). Do not compare similarly named examinations as though they were the same object.
• Record Transfer Syntax UID (0002,0010) from DICOM file meta information. For a network transaction, ask for the accepted transfer syntax from the association logs; the UID is not an ordinary dataset tag carried identically on the wire.
• Record Photometric Interpretation, Samples per Pixel, Planar Configuration where applicable, Bits Allocated, Bits Stored and any palette descriptors/data. Include optional Ultrasound Color Data Present as present/value or absent.
• Record exporter and receiving software builds, routes, retrieval methods, visible errors and both viewers’ exact versions. Keep a record of approved handling and any transformation.

2. Compare compression history and provenance. Lossy Image Compression (0028,2110) records a history of lossy compression; once set to 01 it must remain so after decompression. A present 00, an absent flag, an unchanged UID or an uncompressed transfer syntax is not proof that no alteration occurred. Compare the source and received representations and transformation logs. The lossy-compression history guide covers the detailed ledger. ⁷

3. Make the byte comparison answer a defined question. Matching cryptographic hashes of two files support byte-for-byte identity of those captured files. Different hashes do not prove clinically meaningful change: file metadata, serialization or a permitted representation conversion can differ. A pixel-level comparison across syntaxes needs qualified decoding, frame matching and explicit treatment of color conversion. No such experiment was performed for this article.

4. Compare controlled rendering. Open the same captured object and frame in two facility-approved viewers, recording color settings and relevant display conditions. Different results isolate a rendering difference for investigation, not an automatically correct reference. The 2022 pydicom discussion is a useful warning: a header can disagree with the data, and software may compensate differently. Do not certify the scanner or the clinical adequacy of an examination from a two-viewer comparison alone. ³

Keep transformation separate from correction. Displaying YBR as RGB is normal rendering. Color-space arithmetic may involve rounding; saving through a lossy codec can introduce additional loss. Decompression alone does not add another lossy compression generation. If a supplier proposes correcting a malformed object, require an authorized workflow with the original retained, the change documented and the resulting object validated. Do not silently overwrite the diagnostic archive.

When should the evidence go to PACS support or scanner service?

Use the matrix below to assign an investigation owner. Its rows are possible evidence patterns, not validated diagnoses or compulsory repair steps. A fault beginning after an upgrade is a useful time correlation; it does not establish that the upgrade caused it. Export configuration changes and software recovery need the supplier’s supported procedure and facility change control.

An on-console discrepancy broadens the investigation beyond export rendering, but it still does not prove a beamformer, probe or processing-board failure. Record whether the finding appears in a saved frame as well as the current display. Follow qualified, model-specific assessment rather than resetting presets, reloading software or swapping parts from this article. Preserve existing settings and records before any authorized intervention.

If the visible color or scale is disputed, the clinical team must decide whether the affected representation is fit for the intended clinical use and what alternative is needed under facility policy. A familiar red/blue appearance is not a return-to-service criterion. Where a hardware symptom survives the export checks, the probe-versus-scanner artifact framework provides a separate service discussion; it does not validate the exported object.

Where Rongtao fits—and where it does not

Rongtao Medical provides independent multi-brand ultrasound parts and repair services. When qualified assessment points to a system, board or probe issue, the evidence pack helps define a useful enquiry. Viewer decoding and hospital archive administration belong with the facility’s imaging informatics team and its software suppliers. OEM names identify equipment; they do not imply authorization or endorsement. ⁸

For board repair, current company service information describes a typical 5–8 business-day turnaround, a typically 90-day warranty and 48-hour real-machine testing. Confirm applicable scope, lead time, acceptance evidence and terms for the actual item. A real-machine test does not guarantee clinical reliability or replace the receiving facility’s acceptance and release process. ⁸

Use ultrasound services for the relevant service scope and contact Rongtao with the bounded evidence pack. Keep the original studies within the facility’s approved handling process. For support samples, arrange a lawful, secure transfer and consider a representative non-patient test object where it reproduces the issue.

Frequently asked questions

Does wrong color mean the velocity data changed?
The appearance alone cannot answer that. A renderer may misinterpret unchanged samples, or the export/transfer may have changed the representation. Displayed pixel colors also do not establish that independent numerical velocity data was delivered. Preserve the object and compare the relevant data path before making either claim.

Can I just change Photometric Interpretation to RGB?
No tag-only fix is justified by a familiar-looking result. The declaration must match the encoded samples and transfer syntax. A supported decoder may produce RGB for display without rewriting the stored object. Suspected malformed exports require supplier review and a documented correction process. ^{2 3}

Is a grayscale RGB image automatically broken?
No. Equal RGB components can legitimately yield gray. Ultrasound Color Data Present is optional; if absent, it does not settle whether the intended frame included color. Compare the selected capture and source settings rather than inferring a failed color-processing board. ¹

Do two viewers agreeing prove the file is valid?
No. They may share a decoder, apply similar corrections or accept the same inconsistency. Agreement is one observation. A conformance investigation needs the object type, encoding, relevant supplier declarations and explicit validation results.

What should trigger a fresh review?
Recheck after exporter, viewer, decoder, PACS or routing changes; when a new object type or transfer syntax enters the workflow; or when the relevant standard or supplier conformance statement changes. Record the successful test configuration and its limits so that an old support-thread workaround does not become an unexamined fleet-wide rule.