What happens to your images and reports after the read?

Why are the images and the report stored separately?

Because they are different kinds of data handled by different systems. The images are DICOM objects held in an archive built for large binary files. The report is a text document that lives in the radiology information system or the electronic record. Each has its own storage, its own retention rule, and often its own department.

A chest CT can run to thousands of individual images across multiple series and reconstructions. The signed report fits on a page. From a storage point of view those two things have almost nothing in common, so organizations put them in different places. Images go to the PACS or a vendor neutral archive. The report goes to the radiology information system, the electronic health record, or both. Different systems, different backup schedules, different administrators, and in most organizations different budget lines. The practical consequence shows up at the release desk. A patient asks for a copy of the scan, the release clerk pulls what the record system hands her, and the envelope arrives holding two pages of report and nothing else. Nobody withheld anything. The request was answered literally.

What ties the two records together is a small set of identifiers: the accession number, the study instance UID, the medical record number, and the order. When those match, a report opens next to its images in one click. When they do not match, both records still exist and neither one can find the other. That happens more often than people admit after a merge of duplicate patient records, when an accession number is reissued, or when a study was performed under a temporary trauma identifier that nobody ever reconciled. Health information management staff usually end up finding it by hand.

Who is the custodian of the imaging record?

The facility that performed the examination. It holds the archive, the order, and the consent, and its health information office answers requests. A remote reading service produces the professional interpretation and works inside the facility's systems under contract. It is generally not the custodian and is usually in no position to release records on the facility's behalf.

Custodian is an operational role before it is a legal term. The custodian is the party that can go into its own system, produce what is there, and attest that the copy reproduces the record as kept in the ordinary course of business. That takes holding the archive, the order, the consent, the audit trail, and the policy governing all of it. In imaging, that party is almost always the facility that put the patient on the table: a hospital, an outpatient imaging center, an emergency department. Everyone else in the chain holds a piece and can speak only for that piece.

A teleradiology practice sits on the other side of the contract. It is engaged to interpret studies, it works inside systems the facility controls or has authorized, and whatever copy it retains is normally a working copy governed by the service agreement rather than the legal record of the encounter. Attorneys still subpoena the reading group sometimes, usually because the radiologist's name is the only one printed on the report. That group can rarely produce a certified image set, because it did not acquire the study and cannot speak for what the facility's archive holds today. Sending the request to the facility is faster and produces a copy that is far harder to challenge on authentication grounds.

Mobile and contracted services complicate the picture. When a mobile ultrasound unit performs a study at a long term care home, or a contracted imaging company operates equipment inside a hospital, the archive may sit with either party and the records agreement decides which one. The report letterhead will not tell you. It usually carries the reading group and the ordering site, and says nothing about who stores the pixels. The nursing home may hold the order and the consent while the study itself never touched its systems. Ask who holds the archive before you file the request rather than after, because a request sent to the wrong party still costs you the turnaround before it comes back refused.

How long are images and reports actually kept?

It depends on where the study was performed and which policy applies. Retention is driven by the law of the jurisdiction, the facility's written policy, accreditation and licensure conditions, contract and payer terms, and any litigation hold in force. Facilities plan to whichever rule runs longest. Anyone who needs a firm date should ask that facility in writing.

There is no single retention number, and anyone who gives you one without asking where the study was performed is guessing. The sources stack on top of each other and the longest one governs. Studies on minors are commonly held well past the general period, measured from the age of majority rather than from the date of the examination. Breast imaging in the United States sits under its own federal framework, and departments usually run it on a stricter schedule than everything else they hold. The correct answer to the question is rarely a number. It is the name of the office you have to ask.

Retained and retrievable are not the same thing. Older studies often get tiered off primary storage onto cheaper media or a cloud archive, and pulling one back can take a day or several. Some facilities still keep an offline copy that a technician has to mount, and that technician may work one shift a week. If you are working to a deadline, ask two questions instead of one: is the study still inside the retention period, and how long does retrieval actually take. The second answer is what decides whether the disc reaches you before your hearing.

Policy also has an end. Facilities run purge processes, and once a study is destroyed under a documented schedule it is gone, including from backups in most designs. If litigation is coming, the hold has to reach the imaging archive by name. A hold placed on the chart in the electronic record does not automatically stop a purge job running against the PACS. Those are separate systems, and the people who administer them are separate too. The hold notice that works names the archive, the modality, the date range, and the person in imaging informatics who has to acknowledge it. The one that fails goes to the medical records office alone and is filed as received.

How does a patient or authorized representative request copies?

In writing, to the health information management or release of information office of the facility that performed the study. The request should identify the patient, the examination, the date or date range, and state plainly what is wanted: images, report, or both. A representative also has to prove authority. Most facilities charge for media.

Write to the facility, not to the radiologist and not to a reading service. Most facilities keep an office for this, called health information management, medical records, or release of information depending on the building. Put it in writing even where a phone call is accepted, because the written version is what gets logged, tracked, and answered. Then say exactly what you want. This is where most requests fail. Someone writes “please send my records for the MRI,” the office mails the signed report, and the requester does not discover that the images never came until weeks have gone by. Ask for the DICOM study and the report, in those words.

Authority is the second failure point. A patient signing for herself needs identification. Anyone else needs a document that actually covers health information: a parent of a minor who is still a minor, a guardian with the court order, an executor with letters, an agent under a health care power of attorney rather than a general financial one. Authorizations also expire. Attorneys regularly send one the client signed months earlier for a different purpose, the release office rejects it, and two weeks disappear. Check the date and the scope before it goes in the mail.

Fees and timelines are set by policy and by the law that applies to that facility. Ask for both in writing at the start. Ask also how the copy will be delivered, because a facility that only issues discs and an office with no optical drive have a problem that neither of them will discover until the envelope arrives. What is described here is how release offices generally work. It is general information, not legal advice, and it does not establish what the law requires of any particular facility or of you.

• The patient's full legal name at the time of the study, and date of birth
• The facility and, where known, the specific location where the examination was performed
• Modality and body part, with the date or a bounded date range
• What is being requested in explicit terms: the DICOM image set, the signed report, or both
• A delivery format you can actually open, and the address or portal for delivery
• A signed, current authorization, plus proof of authority if you are not the patient

What does a legally usable copy look like?

It is the DICOM study as it left the archive, with identifiers and metadata intact, delivered on media or through a transfer link, together with the signed report. Anything flattened into a JPEG, a screenshot, or a photograph of a monitor loses the measurements, display settings, and provenance that make the study reviewable.

A copy is usable when another radiologist can open it and work. That means DICOM objects with the header intact: patient and study identifiers, modality, series structure, slice thickness, acquisition parameters, and pixel data at full bit depth. A reviewer needs to scroll the series, change window and level to see soft tissue and then bone, measure a lesion, and check the technique that produced the images. All of that lives in the file, not on the screen. Completeness matters just as much. A study delivered without its thin slices, its reformats, or its post contrast series is technically DICOM and still not enough to review, and the reviewer often cannot tell what is missing without the original series list to compare against.

A photograph of a monitor gives you one frame, at one window setting, with glare and moire, and no way to prove where it came from. A JPEG export is better and is still not the study, because it has thrown away the bit depth and the header. For a second opinion neither is enough, and a reviewing radiologist will usually say so in writing rather than interpret them. For a legal file the problem is provenance. A screen photo cannot be authenticated as a record kept in the ordinary course of business.

Media is uneven. Some discs carry an embedded viewer and some do not, and some are written in a layout a hospital PACS will refuse to import without manual work. Others bury the report inside a viewer that only runs on one operating system, which is a problem the day the file has to be opened on something else. At larger organizations transfer links have mostly replaced discs, and those links expire on a timer nobody warns you about. When you request, ask three things: whether the media is DICOM conformant, whether the signed report is included as a separate document, and, if the copy is going into a legal proceeding, whether the facility provides a custodian certification with it.

• The complete DICOM study, every series, with headers and identifiers unaltered
• Pixel data at original bit depth, not a screen capture or a compressed export
• A viewer, or written confirmation that the media is conformant and imports directly
• The signed report as a separate document, together with any addendum
• A statement of what was included and what was not, such as priors held at another organization
• A custodian certification when the copy is going into a legal file

What happens to old studies when a facility changes PACS vendors?

They have to be migrated, and migration is where records quietly go missing. Some facilities move everything, some move a recent window and leave the rest on the old archive, and some keep the legacy system powered on for lookups only. Ask which archive holds the year you need before assuming the study is gone.

A migration reads studies out of the old archive and writes them into the new one, and the whole job depends on the tags surviving the trip. Most of them do. The ones that do not are the odd cases: studies indexed under a medical record number retired in a duplicate merge, accession numbers in a format the new system rewrites, secondary capture objects with nonstandard tags, studies from a modality decommissioned before the migration was planned. At the end the counts reconcile closely enough that the project is declared finished. Nobody looks again for two years, until somebody requests one of the failures and the archive returns nothing.

The second problem is ownership. On paper, retention belongs to health information management. The archive belongs to IT. When the question is how long the old system stays powered on, the two teams usually find they have different answers and that no budget line owns the license renewal. What follows is a legacy PACS kept alive informally, off support, on hardware nobody will patch, holding the only copy of several years of studies. That is a records risk, and calling it a safety net does not change what it is.

For a facility, the fix is unglamorous: a written migration plan, reconciliation counts by year and by modality instead of one total, a sample retrieval test pulled from the oldest years before the old system is switched off, and a decommissioning date signed by whoever answers for retention. For a requester, the useful move is to name the year. If the study predates a conversion, the facility's own staff may have to look in a different system to find it, and saying so up front saves a wasted round trip.

Sources and scope

• National Electrical Manufacturers Association, DICOM Standard, including media storage and interchange
• American College of Radiology, practice parameters and technical standards for medical imaging
• American Health Information Management Association, professional guidance on health record retention and release of information
• Health Level Seven International, standards for exchanging clinical documents and results
• Radiological Society of North America, professional education on imaging informatics and image sharing