Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Home·Academy·Case report·Radiology case reports

How to Write a High-Impact Radiology Case Report for Physicians

A radiology case report differs from other case reports in one respect that governs everything else. Its evidence is images, and an image can identify the patient in three ways at once. Information sits in the file header, in text burned into the pixels, and in the anatomy itself. Deleting the header addresses the first of those. It addresses neither of the other two.

What you will learn
  • The three separate places a radiological image carries identifying information
  • What two radiology journals actually require of figures, consent and manipulation
  • Which adjustments to an image are permitted and which must be disclosed
What to unlearn
  • “Stripping the DICOM header de-identifies the image.” It leaves burned-in pixel text untouched. It also does nothing about anatomy from which a face can be reconstructed.
  • “Masking the eyes is enough.” One journal's policy states directly that simple masking of eyes in photographs does not preserve anonymity.

What this article covers

Guidance on case reports is abundant and almost all of it concerns text. It explains how to choose a case, how to structure the narrative, and which checklist to follow. None of that is wrong, and this article does not repeat it.

What general guidance does not cover is the part specific to radiology. The figures are not illustrations of the argument; they are the argument. That changes the obligations attached to them. It also introduces failure modes with no counterpart in a text-only report.

The article therefore concentrates on images. It covers where identifying information hides, what the evidence says about removing it, what two radiology journals require, and which adjustments are permitted. The structure and reporting checklist for case reports generally are covered in the companion article linked below.

Where this sits among the related articles

The reporting checklist mapped section by section, together with consent and de-identification as general requirements, is covered in writing an effective medical case report. The order in which a case is told is covered in structuring a case presentation. Reuse and attribution of previously published figures are covered in avoiding plagiarism and establishing originality. This article is about the images.

The three places an identifier hides

A radiological study is a file, a picture and a body part at the same time. Each of those carries identifying information by a different route, and each needs a different remedy.

The file header

DICOM files carry structured metadata. A de-identification study working from the standard's own confidentiality profile targets seven core tags.1 They are patient name, patient identifier, date of birth, study date, accession number, institution name and referring physician name.1 It extends that to further demographic and acquisition tags, including sex, age, study time and study identifier.1

Two of those deserve notice because authors rarely think of them as identifying. The study date is one. Under the safe harbour method of de-identification, all date elements directly related to an individual must be removed except the year.5 The institution name is the other, since it narrows the population from which the patient could be drawn.

The pixels

Text burned into the image itself is a separate problem, and the de-identification study calls it the major challenge.1 The optional DICOM attribute that marks an image as containing burned-in annotation “is not always applied”.1 A file may therefore carry such text without declaring it. Vendor-specific private tags present a related difficulty, holding information the standard's specification does not cover.1

Burned-in content is not limited to patient names. A case repository's publishing standard asks for images “ideally devoid of ALL text overlay”.6 It extends that to watermarks, the contributor's name, the institution's name and logo, and the branding of the archiving system or viewer.6 It asks that manufacturer logos burned into images, particularly in ultrasound, be removed where possible.6

The anatomy

The third location has no software remedy of the usual kind, because the identifying information is the scan. Cross-sectional imaging of the head contains the surface of the face, and a surface can be reconstructed and matched. The evidence for that is set out in the next section but one.

Regulation anticipates part of this. The safe harbour list of identifiers includes full-face photographs “and any comparable images” as item 15, and biometric identifiers as item 13.5 A further item covers any other unique identifying number, characteristic or code.5 Whether a reconstructable head scan falls under the comparable-images wording is a question for the institution and the journal. It is not one this article can settle.

Why the three need separate treatment

A script that clears header tags leaves pixel text intact, because the two are stored differently. A tool that redacts pixel text leaves the anatomy intact. The anatomy is the diagnostic content, and removing it would destroy the figure. No single operation addresses all three, which is why de-identification of imaging is a sequence of checks rather than a button. The sections below report what has been measured about each.

Figures: protecting patient identity in imaging

The two figures below support the consent and de-identification steps of a radiology case report. Figure 1 maps where identifying information sits in a radiological study and what removes it from each location. Figure 2 summarises the evidence that facial features reconstructed from MRI can be matched to photographs. Use them as a checklist before images are submitted.

Figure 1 · Three locations, three remedies

What each part of a radiological study carries, and what removes it.

ONE STUDY, THREE SEPARATE CARRIERS1 · THE FILE HEADERName, identifier, date of birth, study date, accession number,institution, referring physician, and vendor-specific private tagsCLEARING TAGS WORKS2 · THE PIXELSBurned-in annotation, which the attribute that should flag itdoes not always declare; watermarks, institution and vendor logosNEEDS DETECTION3 · THE ANATOMYThe surface of the face, carried in any cross-sectional studyof the head, from which a likeness can be reconstructedNEITHER REMEDY APPLIESA script that clears the header leaves the pixels. A tool that redactsthe pixels leaves the anatomy, because the anatomy is the figure.No single operation reaches all three.

What to look for. The three rows are not degrees of the same problem. They are different problems stored in different ways, and the remedy for each is useless against the other two. The third has no software fix of the usual kind. Removing the identifying content would remove the diagnostic content with it.

The header fields listed are those targeted by a published de-identification pipeline working from the standard's confidentiality profile. The pixel row describes what that same work identifies as the harder case.

Figure 2 · Matching face reconstructions to photographs

84 research participants; 3D faces reconstructed from MRI, matched against photographs by commercial face-recognition software.

PARTICIPANTS CORRECTLY MATCHED, OF 84Top match70 · 83%Among top five80021426384WHERE THE IDENTIFYING CONTENT WASNot in the header and not in the pixel text, but in the surface ofthe face, which a de-identification step does not touch.

What to look for. The top-match figure is the one that matters, because an attacker does not need five guesses to cause harm. Ninety-five per cent within the top five says the signal is not marginal. Both results come from a single study of 84 participants, which is small, and the software was a commercial product rather than a specialised tool.

Each participant's scan produced a reconstructed face rendered as ten two-dimensional images, matched against five photographs of each person, with scores summed across the five.

Re-identification from the anatomy

The claim that a head scan can identify its subject is not a precaution invented by ethics committees. It has been tested, and the result is specific enough to act on.

The method of the test

Eighty-four research participants took part.2 Each participant's magnetic resonance scan was used to reconstruct a three-dimensional face.2 Ten two-dimensional images were generated from it.2 Five real photographs of each person were then matched against the set of 84 reconstructions using publicly available face-recognition software.2 Scores were summed across each person's five photographs.2

The result

The authors state it directly. “For 70 of the 84 participants (83%), the software chose the correct MRI scan as the most likely match for their photos.”2 They add that “The correct MRI scan was among the top five choices for 80 of 84 participants (95%).”2

The first of those is the operative figure, because an attempt at identification does not need five guesses to succeed. The second establishes that the signal is not marginal. A near-miss still places the right person in a very short list.

The reason a header fix does not help

Nothing in that procedure touched metadata. The reconstruction used the scan's own content, which is the information the figure exists to show. A routine that cleared every tag in the file would have left the result unchanged. The surface of the face is not stored in a tag.

The regulatory wording is relevant without being decisive. The safe harbour list names full-face photographs “and any comparable images”, lists biometric identifiers separately, and closes with any other unique identifying characteristic.5 Whether a reconstructable head scan is a comparable image under that wording is for an institution and a journal to determine. This article does not determine it. The measured match rate is what an author needs in order to raise the question with them.

The limits of a single study

Eighty-four participants is a small sample, and the work used one commercial face-recognition product rather than a specialised tool. It concerns magnetic resonance imaging of the head, and this article does not extend the finding to other modalities or other body regions. What it establishes is that the risk is real and measurable where a face is in the field of view. That is the case a neuroradiology report most often involves.

The consequence for a published case

A published case report is more exposed than a research archive, not less. The scan sits on an open web page rather than behind a data-access agreement. The accompanying text supplies clinical detail that narrows the population further. Where a case turns on head imaging, defacing or cropping the reconstructable surface should be considered before submission rather than after a reader raises it.

The header and the pixels after de-identification

The first two locations do have software remedies, and the question is how completely they work. One published pipeline reports its own performance in enough detail to answer that.

The scope of the pipeline

The work combined deep learning and language processing to remove protected information from both metadata and burned-in pixel text.1 It went beyond the seven core tags of the standard's confidentiality profile.1 Pattern-based checks were applied across free-text value representations, to catch information held in non-standard or vendor-specific private tags that the specification does not cover.1

One detail from the test material is worth carrying. Headers in a public phantom dataset contained nine such tags per file, including patient name, patient identifier and institution name.1 Pixel-level text there was confined to three annotation regions.1 A file does not have to come from a live archive to carry this material.

The residuals it reported

End to end, 227 of 5,933 instances were missed, giving recall of 96.1 per cent and precision of 94.7 per cent.1 The missed cases divided into detection failures at 38 per cent, optical character recognition errors at 27 per cent, and classification errors at 35 per cent.1 Residuals at the masking stage fell from 0.48 per cent to 0.17 per cent once a verification pass was added.1 Fifty-five images, or 5.3 per cent, were flagged for manual review.1

The authors report an earlier figure from the literature for comparison.1 A pipeline specific to ultrasound missed information in 11 per cent of test images.1 Neither number supports an assumption that an automated step finishes the job.

The design principles the authors draw

Three recommendations follow from those residuals, and all three invert the instinct of someone tidying a figure. Masking should be selective and semantic rather than removing every detected string.1 The reasoning is that “missing PHI is more harmful than unnecessarily masking non-PHI content”.1 Black-box masking is preferred to generative reconstruction, described as “a more reliable solution than GAN or diffusion-based reconstruction methods”.1 A verification pass should run at a deliberately lower threshold, since it is “preferable to flag false positives for brief human review than to miss residual PHI”.1

How to read these performance figures

The information removed in this evaluation was synthetic, embedded into public images rather than drawn from clinical archives.1 Survival rates therefore describe a controlled setup. They are not a measurement of what remains in a real department's files. The direction is still informative. A purpose-built pipeline, measuring itself, reported residuals rather than none.1 It set aside one image in twenty for a person to look at.1

The attribute that cannot be trusted

One finding bears directly on manual checking. The DICOM attribute that marks an image as containing burned-in annotation is optional, and “is not always applied”.1 An author cannot infer from its absence that an image is clean. The only reliable test is to look at the image at full size. That is also the step that catches a logo or a viewer watermark.

The rules two radiology journals apply

Journal policy is where these obligations become concrete, and two radiology journals set out requirements an author can work against. They agree on the substance and differ in emphasis, which is useful because the overlap marks what is settled.

Consent and identifiability

One policy requires authors to certify that informed consent was obtained from human subjects.3 All potentially identifying information must be removed before submission.3 That extends to images, charts, graphs, tables and text, naming patient likenesses, identification numbers, names and initials.3 Where an identifiable image is essential, a journal-specific consent form is required, and the individual must be offered the chance to see the manuscript.3

The same policy contains the sentence most worth quoting on this subject. “The simple masking of eyes in photographs is not sufficient to preserve anonymity.”3 For photographs or three-dimensional renderings it allows that a cropped or adequately masked image excluding identifiable features may be acceptable.3

The second journal requires appropriate consents, permissions and releases where an article includes case details, personal information or images of patients.4 Personal details must be removed from every part of the article, including all illustrations and videos.4 The exception is written permission from the patient or next of kin.4 Both policies keep the signed forms with the author rather than the journal.4 The second states that copies should not be provided unless specifically requested in exceptional circumstances.4

The permitted and prohibited adjustments

The second journal states the manipulation rule in a single line. “No specific feature within an image may be enhanced, obscured, moved, removed, or introduced.”4 Adjustments to brightness, contrast and colour balance are acceptable provided they do not remove information.4 Nonlinear adjustments such as changes to gamma settings “must be disclosed in the figure legend”.4

The first journal reaches the same place by a different route. It makes authors responsible for the integrity of their digital image files.3 Where images have been enhanced or manipulated, “the specific alterations must be disclosed at the time of submission”.3 It asks the corresponding author to keep a complete set of digital image files for future reference.3

The rule most often broken by accident

Obscuring a feature is prohibited in the same breath as introducing one, and that is the half authors overlook. Masking a burned-in identifier that sits over anatomy falls under the same prohibition as adding something that was not there, and so does cropping to hide an incidental finding. The remedy is not to leave the identifier in place. It is to mask it, disclose the mask, and keep the original file so that the editor can see what was covered.3

Format, legends and credit

Specifications are given rather than implied. Halftone photographs are required at a minimum of 300 dots per inch.4 Bitmapped line drawings require a minimum of 1000, and combined line and halftone images a minimum of 500.4 Screen-optimised formats are not accepted.4 Captions are supplied separately rather than attached to the figure, and all symbols and abbreviations must be explained.4 A legend for an imaging figure is expected to carry the patient's age and sex and the main clinical topic.4

One rule concerns authorship rather than figures, and it is specific to this field. “The sole contribution of radiologic images or other illustrations does not qualify a person as an author.”3 Supplying the case images earns acknowledgement, not a byline. Previously published figures must be identified in the legend as reprinted or adapted with permission.3 Unpublished borrowed or commissioned artwork requires a signed letter of consent from its creators at initial submission.3

The repository standard and where it stops

Many radiology cases are published to teaching repositories rather than to journals. The publishing standard of one such repository is worth reading alongside journal policy, because its operational detail is more specific and its gaps are instructive.

The requirements for the image itself

The standard asks that submitted cases be completely de-identified.6 It acknowledges that rarity, presentation, demographics and the size of a practice can still make identification possible.6 On the image itself it is more exacting than most journal policies. Images should be ideally devoid of all text overlay, and the stated best approach is to have zero text overlays.6

The prohibition extends well beyond patient details. No watermarks, logos or other graphics should appear on a clinical image.6 That includes the contributor's own name, the institution's name or logo, and the branding of the archiving system or viewer.6 Manufacturer logos burned into images, particularly in ultrasound, should be removed where possible.6 The enforcement is blunt: a case uploaded with patient information will be deleted immediately.6

The annotation rule worth adopting anywhere

Where arrows or labels are wanted, the standard asks for a separate annotated version of a single image rather than marks applied to the primary images.6 That practice solves a problem journal policy creates but does not resolve. An annotated image has had a feature introduced, which the manipulation rule prohibits. An unannotated original beside a clearly labelled annotated copy satisfies both the teaching purpose and the integrity requirement.

Where the standard stops short

Two silences are worth noting, because a reader following this guidance alone would not be covered. The standard does not mandate patient consent for all contributions.6 It holds instead that confidentiality must be maintained and local policy followed, and asks contributors to confirm that their institution permits publication.6 It urges obtaining informed consent where there is doubt.6 A journal submission is governed by the journal's policy, which in both cases examined here does require consent.34

The second silence is larger. The standard gives no instruction for stripping metadata, and it does not mention facial features or three-dimensional reconstructions at all.6 Given the match rates reported for reconstructed faces, that is a gap an author has to close themselves.2

Why repository rules are stricter about overlays

A teaching repository receives images exported directly from a workstation, where overlays are the normal state rather than the exception. Journal policy assumes a figure prepared for publication; a repository assumes a screenshot. The stricter wording reflects where the images come from. An author exporting from a viewer is in the second position whichever destination the case is going to.

Image practice that survives review

The requirements above converge on a short sequence. It is ordered so that each step catches what the previous one cannot, which is the only arrangement that works when three different carriers are involved.

Export, then strip, then look

Metadata is cleared first, because it is the only part a tool reliably finishes. That means the core identifying tags and also the ones authors forget, such as the study date and the institution name.1 Private and vendor-specific tags need separate attention. They hold material the standard's profile does not cover.1

The image is then examined at full size, by eye. The attribute that should declare burned-in annotation is optional and is not always applied, so its absence proves nothing.1 This pass looks for more than a name. It looks for the accession number in a corner, the institution's logo, the viewer's watermark, and a manufacturer's mark burned into an ultrasound frame.6

Decide the anatomy question explicitly

Where the case rests on cross-sectional head imaging, the face is in the dataset and a reconstruction can be matched to a photograph in most attempts.2 The options are to deface the surface, to crop so that it is not included, or to obtain consent that covers publication of the image. Journal policy allows a cropped or adequately masked image excluding identifiable features.3 What is not an option is to assume the question does not arise because the header is clean.

Keep the originals and disclose the changes

The original unaltered files are retained. One policy asks for that directly, requiring a complete set of digital image files for future reference.3 Any adjustment beyond brightness, contrast and colour balance is disclosed, and a nonlinear change such as gamma goes in the figure legend.4 Masking applied to remove an identifier is an alteration like any other, and is disclosed at submission rather than left for an editor to notice.3

Annotations live on a duplicate. The primary figure stays unmarked, and a separate labelled version carries the arrows.6 That keeps the teaching value without introducing a feature into the image of record.

Settle consent and credit before submission

Consent is obtained in the form the target journal specifies, and the signed document stays with the author rather than being sent to the journal.34 Where an identifiable image is essential, one policy additionally requires that the individual be offered the chance to see the manuscript.3

Credit is settled at the same point. A colleague who supplied the images and nothing else is acknowledged rather than listed as an author.3 The sole contribution of radiologic images does not qualify a person for authorship.3 Any previously published figure is identified in its legend as reprinted or adapted with permission.3

The check that catches most of it

Open each final figure at full size on a screen, alone, after every other step is finished. Automated de-identification reported residuals even when measuring itself, and set aside roughly one image in twenty for human review.1 The eye catches what a detector misses. It is also the only step that sees the figure the way a reader will.

Take-home message

A radiological study carries identifying information in three places, and each needs a different remedy. The file header holds structured metadata including the study date and the institution name, and clearing tags removes it. The pixels may hold burned-in text, watermarks and logos, and the attribute that should declare them is optional and not always applied. The anatomy holds the surface of the face in any cross-sectional study of the head, and no de-identification step touches it. Reconstructed faces were matched to photographs for 70 of 84 participants. Journal policy requires that identifying information be removed before submission. It requires that no feature within an image be enhanced, obscured, moved, removed or introduced, and that any alteration be disclosed. Supplying the images alone does not qualify a colleague for authorship.

How to apply this to your work

In order, from export to submission
  1. Clear the metadata first, including the study date and institution name, and check vendor-specific private tags separately.
  2. Open every image at full size and look at it, since the attribute that should flag burned-in annotation is optional and often absent.
  3. Remove overlays beyond patient details, including viewer watermarks, institutional logos and manufacturer marks.
  4. Decide the face question explicitly on any head study, choosing between defacing, cropping or consent that covers image publication.
  5. Keep the original unaltered files, which one policy requires as a complete set for future reference.
  6. Put annotations on a duplicate, leaving the primary figure unmarked.
  7. Disclose every alteration at submission, and put nonlinear adjustments such as gamma changes in the legend.
  8. Settle consent and credit before you write, remembering that supplying images alone earns acknowledgement rather than authorship.

Common questions

Does removing the DICOM header de-identify an image

No, it addresses structured metadata only. Text burned into the pixels is stored differently and survives untouched, and the optional attribute that marks an image as containing such text is not always applied.1 The anatomy is a third carrier that no header operation reaches.2

Can a brain scan identify the patient

In most attempts, on the available evidence. Faces were reconstructed from the magnetic resonance scans of 84 participants and matched against photographs using publicly available software. “For 70 of the 84 participants (83%), the software chose the correct MRI scan as the most likely match for their photos.”2 The correct scan was in the top five for 80 of 84.2

Is masking the eyes enough

It is not, and one radiology journal says so directly: “The simple masking of eyes in photographs is not sufficient to preserve anonymity.”3 The same policy allows that a cropped or adequately masked image excluding identifiable features may be acceptable for photographs or three-dimensional renderings.3

Which adjustments to a radiological image are permitted

Brightness, contrast and colour balance, provided they do not remove information.4 The governing rule is that “no specific feature within an image may be enhanced, obscured, moved, removed, or introduced”, and nonlinear adjustments such as gamma changes “must be disclosed in the figure legend”.4 Enhancement or manipulation must be disclosed at submission.3

Is the study date really an identifier

Under the safe harbour method, all date elements directly related to an individual must be removed except the year.5 The study date is among the core tags a published de-identification pipeline targets, alongside patient name, identifier, date of birth, accession number, institution and referring physician.1

How well do automated de-identification tools perform

Well, but not completely. One pipeline missed 227 of 5,933 instances, a recall of 96.1 per cent, and flagged 5.3 per cent of images for manual review.1 The same work cites an earlier ultrasound-specific pipeline that missed information in 11 per cent of test images.1 The material used in that evaluation was synthetic rather than clinical.1

What has to come off an image besides patient details

Watermarks, logos and other graphics, including the contributor's own name, the institution's name or logo, and the branding of the archiving system or viewer.6 Manufacturer logos burned into images, particularly in ultrasound, should be removed where possible.6

Does supplying the images make a colleague an author

It does not qualify them. One radiology journal states that “the sole contribution of radiologic images or other illustrations does not qualify a person as an author”.3 Acknowledgement is the appropriate credit. Previously published figures must be identified in the legend as reprinted or adapted with permission.3

Declared interest: Pubrica provides medical writing and publication support commercially. The re-identification result comes from a single study of 84 participants using one commercial face-recognition product, and concerns magnetic resonance imaging of the head. It is not extended here to other modalities or body regions. The de-identification performance figures were measured on synthetic information embedded into public images rather than on clinical archives. They describe a controlled setup rather than a department's files. Journal requirements described here are those of the two journals named and do not substitute for the instructions of the journal an author is submitting to. Whether a reconstructable head scan falls within the regulatory wording on comparable images is a determination for an institution and a journal, and no view is offered on it.

References and sources

  1. Sethulekshmi R, Pillai MJ, Ahammed N, Ramasubbareddy S, Cho YY. Semantic de-identification of burned-in PHI in DICOM medical images: a deep learning–NLP pipeline validated on clinical and phantom TMM datasets. Frontiers in Computer Science. 2026;8:1838941. doi:10.3389/fcomp.2026.1838941View the full text
  2. Schwarz CG, Petersen RC, Jack CR Jr, et al. Identification of anonymous MRI research participants with face-recognition software. New England Journal of Medicine. 2019;381(17):1684–1686. doi:10.1056/NEJMc1908881View the correspondence
  3. Radiological Society of North America. RadioGraphics publication information for authors: editorial policies on patient privacy, image integrity, figures and authorship.View the policies
  4. Radiología. Instructions for authors: article types, figure preparation, image manipulation and patient consent. Sociedad Española de Radiología Médica, published by Elsevier.View the instructions
  5. American Heart Association. HIPAA compliance for open data policy. July 2026.View the policy
  6. Radiopaedia editorial board. Case publishing guidelines.View the guidelines

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 22). The Radiology Case Report: What the Images Oblige You To Do. Pubrica. https://pubrica.com/academy/case-report/radiology-case-report-guide/
Pubrica Academy. “The Radiology Case Report: What the Images Oblige You To Do.” Pubrica, 22 Oct. 2026, pubrica.com/academy/case-report/radiology-case-report-guide/.
Pubrica Academy (2026) The Radiology Case Report: What the Images Oblige You To Do. Available at: https://pubrica.com/academy/case-report/radiology-case-report-guide/ (Accessed: 22 October 2026).
Pubrica Academy. “The Radiology Case Report: What the Images Oblige You To Do.” Pubrica. October 22, 2026. https://pubrica.com/academy/case-report/radiology-case-report-guide/.
Pubrica Academy, ‘The Radiology Case Report: What the Images Oblige You To Do’ (Pubrica, 22 October 2026) <https://pubrica.com/academy/case-report/radiology-case-report-guide/> accessed 22 October 2026.
Pubrica Academy. The Radiology Case Report: What the Images Oblige You To Do [Internet]. Pubrica; 2026 Oct 22 [cited 2026 Oct 22]. Available from: https://pubrica.com/academy/case-report/radiology-case-report-guide/
[1] Pubrica Academy, “The Radiology Case Report: What the Images Oblige You To Do,” Pubrica, Oct. 22, 2026. [Online]. Available: https://pubrica.com/academy/case-report/radiology-case-report-guide/
@misc{pubrica_rad_2026, author = {{Pubrica Academy}}, title = {The Radiology Case Report: What the Images Oblige You To Do}, year = {2026}, month = {oct}, day = {22}, howpublished = {\url{https://pubrica.com/academy/case-report/radiology-case-report-guide/}}, note = {Accessed: 2026-10-22} }

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Radiology case report image checklist

Work through these before the figures leave your hands.

THE FILE HEADER
[ ] Patient name, identifier and date of birth cleared
[ ] Study date cleared or reduced to the year
[ ] Accession number cleared
[ ] Institution name and referring physician cleared
[ ] Patient sex, age, study time and study identifier reviewed
[ ] Vendor-specific and private tags inspected separately
[ ] A second file from the same series checked, not just the first

THE PIXELS, CHECKED BY EYE AT FULL SIZE
[ ] Burned-in patient details absent from every corner
[ ] Accession or examination numbers absent
[ ] Institution name and logo absent
[ ] Viewer or archive branding absent
[ ] Manufacturer logo absent, ultrasound especially
[ ] Contributor's own name and institution absent
[ ] Scale bars and laterality markers retained, not stripped by accident

THE ANATOMY
[ ] Head study identified as carrying a reconstructable face
[ ] Decision recorded: deface, crop, or consent covering publication
[ ] Cropping checked against diagnostic content before it is applied
[ ] Three-dimensional renderings reviewed for identifiable features

CONSENT AND PERMISSION
[ ] Consent obtained in the form the target journal specifies
[ ] Signed form retained by the author, not sent to the journal
[ ] Institutional permission to publish confirmed
[ ] Where an identifiable image is essential, patient offered the manuscript
[ ] Permission held for any previously published figure

IMAGE INTEGRITY
[ ] Original unaltered files retained as a complete set
[ ] No feature enhanced, obscured, moved, removed or introduced
[ ] Brightness and contrast adjustments do not remove information
[ ] Nonlinear adjustments disclosed in the figure legend
[ ] Masking applied to remove identifiers disclosed at submission
[ ] Annotations placed on a duplicate, primary figure left unmarked

FIGURES AND CREDIT
[ ] Resolution meets the journal's stated minimum
[ ] Captions supplied separately from the image files
[ ] Legend carries age, sex and the main clinical topic
[ ] All symbols and abbreviations explained
[ ] Image contributors acknowledged rather than listed as authors
[ ] Reprinted or adapted figures identified as such in the legend
PA
Pubrica Academy

Written and maintained by the subject matter experts of Pubrica who do this work. Articles are revised when the standard underneath them moves, and each one carries the date it was last checked against its sources.

How Pubrica helps

We check case figures against the target journal's own image policy before submission, covering metadata, overlays and the integrity rules. We also prepare legends and consent documentation to the form the journal specifies.

Explore

Related reading in the Academy.

Picked from the same sub-topic and category as this piece, newest first.

Find something else

Search the rest of the Academy.

Open with no account. Type what you are working on, or open the full library to filter by category, sub-topic or resource type.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.