CLI privacy audit

This walkthrough uses the command line. See Install first, or follow the separate Desktop privacy audit example. Read Audit modes to distinguish a privacy audit from a dataset comparison and its optional checks.

Install dicomqc and verify dicomqc --version before running the commands below.

Try the built-in demo

Generate three synthetic DICOM files and audit them:

dicomqc demo

Data and expected findings

The files are in dicomqc-demo/dicom/. All names, identifiers, and dates below are invented for the example.

File	Synthetic metadata	Expected findings
raw_phi.dcm	Name Smith^Jane, ID LOCAL123, birth date 19700101, and a private creator tag	1 error for the birth date; 3 warnings for the name, ID, and private tag
pseudonymized.dcm	Name and ID SUBJ001; no birth date or private tags	No findings
private_tags.dcm	Name and ID SUBJ002; no birth date; a private creator tag remains	1 warning for private tags

Expected total: 3 files read, 1 error, 4 warnings. The private-tag warning asks for review; it does not establish that the tag contains identifying data. Likewise, a matching pseudonym format is not proof that a file is safe to share.

The demo intentionally includes findings, so the reported scan exit code is 2. The demo command itself exits 0 when the example was generated correctly.

Open the reports

Open dicomqc-demo/dicomqc/report.html in a browser. It shows the five findings as four grouped issues: the private-tag issue covers two files. The HTML is self-contained and works offline; no server, MultiQC installation, or additional Python packages are needed.

View the HTML report for this scan

Review the HTML report

Each issue shows its recommended action once. Expand View affected files / pairs to see the individual findings; errors appear before warnings and information.

Reading and printing the HTML report

• Charts and audit details contains counts, scope, profile, and any project policy ID and SHA-256 digest. This panel and Search and filter start closed.
• Search matches files, rules, fields, and recommendations. Matching occurrence lists open automatically. Category, severity, and search filters combine.
• Clicking a category bar applies that filter. Reset filters clears filters and collapses lists. Charts always summarize the full audit; category bars count findings, not unique files.
• Expand file lists opens the visible groups together. Rule IDs and standards references are inside these lists.
• Print report includes every finding and supporting detail, even when filtered or collapsed. The browser's print command works too.
• Without JavaScript, all issues remain visible and file lists can still be expanded; filters are unavailable.

Groups share a rule, severity, message, recommendation, and standards references. Repeated fields remain separate occurrences. Narrow screens use labeled rows instead of wide occurrence tables. This demo's HTML does not embed raw records or tag values. The optional vendor inventory explicitly includes observed equipment and creator labels.

The same findings are written to dicomqc-demo/dicomqc/findings.csv; dicomqc-demo/dicomqc/report.json also includes the scan summary and file metadata. See output formats for their structure and report privacy before sharing them.

Rerun the analysis

Audit the generated files again and replace their reports:

dicomqc scan dicomqc-demo/dicom \
  --html dicomqc-demo/dicomqc/report.html \
  --json dicomqc-demo/dicomqc/report.json \
  --csv dicomqc-demo/dicomqc/findings.csv

This scan command returns 2, because the example still contains the same error and warnings. It does not change the DICOM files.

View the same scan in MultiQC

This step is optional. The demo also writes a MultiQC bundle in dicomqc-demo/dicomqc/dicomqc_mqc/. With MultiQC installed, render it with:

multiqc dicomqc-demo/dicomqc --outdir dicomqc-demo --force

Open dicomqc-demo/multiqc_report.html. This displays the same scan findings; MultiQC does not run another DICOM audit.

View the MultiQC overview, status, and findings

The overview shows the same one error and four warnings.

The status table summarizes the audit counts.

The findings table lists the five findings and recommended actions.

To refresh that bundle after another audit, run:

dicomqc scan dicomqc-demo/dicom --multiqc dicomqc-demo/dicomqc/dicomqc_mqc

Then rerun the MultiQC command above. For other examples, follow the complete dataset comparison or project policy walkthrough.

MultiQC controls and bundle files

--multiqc writes custom content, not a finished HTML report. MultiQC displays general statistics, an audit overview, a status table, and individual findings with recommended actions. Technical columns start hidden; use Configure columns to show rule IDs, tags, and value states. Project policy details appear in the overview's expandable notes.

HTML and MultiQC share the same status wording: Errors require attention, Warnings require review, Checks passed, or No readable files. These describe the audit, not permission to share data. MultiQC retains its own tables and export controls; standalone HTML groups repeated findings into issues.

The bundle contains *_mqc.yaml files and an HTML overview. Keep its output directory separate from inputs, policy files, and other report outputs. Existing bundle files are replaced when you rerun the export. Legacy filenames and internal IDs containing release_status remain for compatibility; visible labels say Audit status.

Scan a dataset

Scan a study directory:

dicomqc scan study/ --html report.html --json report.json --csv findings.csv

The scanner recursively attempts regular files under the provided path. DICOM files do not need a .dcm extension.

Exit codes

Code	Meaning
0	No warnings or errors
1	Warnings only
2	Validation errors or fatal scan failures

Default profile

The default profile is research-release-v0.1. It flags direct identifiers, warns when patient identifiers do not match the expected pseudonym format, and lists private tags for review.

Fixing findings

dicomqc does not modify DICOM files. If it reports required changes, apply them with a pseudonymization or anonymization tool, then rerun the scan. See Remediation for examples with external tools.

For the intended multiple sclerosis MRI use case, see MS MRI Workflow.