Glossary

This page gives the one term that the documentation uses for each concept. The terms apply to prose only. An API name stays as it is wherever it names the API, for example anonymize(), date_jitter or config.

The procedure pages and the tutorials use these terms every time. The other pages use them as a guide.

Terms

Term	Meaning	Not used for this concept
de-identify, de-identification	The whole process that removes identifiers.	anonymize, scrub, pseudonymize
anonymize	The metadata step that applies the tag rules: anonymize(), or “the anonymize step”.	—
reversible anonymization	The optional feature that lets the holder of a key recover the original identities. It is the name of the feature and of enable_reversible_anonymization().	reversible de-identification, re-identification
scrub	Remove identifiers from free text with the SCRUB action. Used only for that action.	clean, sanitize
the store	<name>.db and <name>_pixels.bin together. The first mention on a page is “the session store”. “SQLite” names the engine only where the engine matters, for example for WAL or flock.	database, SQLite file, index, persistence file
the sidecar	The append-only file that holds pixel and waveform data, <name>_pixels.bin.	blob file, pixel file
the audit log	The record of actions that the store keeps.	log, trail, history
audit row	One record in the audit log. “A WARNING row” where the level matters.	log entry, audit entry, event
Audit trail	The title of the Audit trail API page. Used only as that name.	—
log line	One line in isocenter.log. A log line is not an audit row: the two are in different places.	log message, entry
finding	A possible identifier that the scan detected (PhiFinding).	hit, detection, match
grade	PASS or REVIEW_REQUIRED. Also a verb: “the report grades the run”.	verdict, status, score
grade condition	One reason why a run is not PASS. How the grade is decided lists them. The report shows them in its Grade Basis.	reason, criterion
PHI status	REMEDIATED, CLEARED, IDENTIFIED or UNSCANNED. “Status” alone means this.	dirty, state, flag
scan, audit	To scan is to read the graph for PHI. The audit is the step that scans, audit(). examine() is a different step.	inspect, examine, check
remediate, remediation	Change a tag as a rule says. The procedures say “remove”, “replace” or “shift” instead.	fix, clean, apply
decline	Not apply a rule, and write a REMEDIATION_DECLINED row. A decline is not an error. IngestSummary.declined is ingest’s count of files whose SOP Instance UID the session already holds, or whose study another patient holds. That use is the API’s own name.	skip, refuse
refuse	Turn an input or a call away with an error or an ERROR row. For example, ingest refuses a file.	reject, deny
withhold	Not write an instance, on purpose.	skip, drop, exclude
data loss	Data that the export could not carry, recorded in a DATA_LOSS row. A verb is fine: “the export drops X”.	“dropped data”, “lost data”
de-identification markers	The elements (0012,0062), (0012,0063) and (0028,0303) that the export writes.	stamps, flags
owner stamp	A Patient, Study or Series field that the export writes into each file.	—
export	Write de-identified copies to a folder. Used as a verb and as the name of the step. “Write” is the act at the level of one file: “the worker writes the file”.	write out, save, output
export folder	The folder that the export writes into.	output directory, destination
ingest	Read source files into the store.	import, load, index
configuration	The YAML policy file and the object that holds it. config appears only in code.	config file, settings
profile	A pinned rule table, such as basic@2026c. Also “privacy profile”.	template, preset
policy	What a PHI status is recorded under: the tag rules, remove_private_tags and CONFIG_VERSION. A policy is not a configuration.	—
redaction zone	A rectangle of pixels that a configuration tells redact() to blank. “Region” and “box” are used only for what OCR or the text detector finds.	region, mask, box
redact	Blank pixels.	mask, black out, burn
burned-in text	Text drawn into the pixels. The hyphen is used when the term comes before a noun. The verb phrase is fine: “text burned in by the modality”.	burnt-in, burned in (before a noun)
date shift, shift	The per-patient offset that moves dates. “Jitter” appears only in the key date_jitter and the action JITTER.	jitter
pseudonym	An ANON_… value.	alias, token
key file	The file that holds the key for reversible anonymization, for example isocenter.key. “Fernet key” is used only where the format matters.	secret
project secret	The per-project HMAC secret that the store holds. It is not the key file, and it is never called a key.	key, salt
identity token	The encrypted identity in (0400,0500). The first mention on a page is “identity token”, not “token” alone.	token (first mention)
cohort	A set of patients that are processed together. “Dataset” is used for a pydicom Dataset and for MIDI-B’s own datasets.	dataset, collection
the graph	Patient → Study → Series → Instance in memory. The first mention on a page is “the in-memory object graph”.	model, tree
instance	The object for one DICOM file. “File” means the bytes on disk.	file (for the object), object
tag, element	A tag is the identifier, gggg,eeee. An element (data element) is the item in a file. “Attribute” is used for attributes and in quotations from the standard.	attribute