# Radiology report: parts, examples and how to review one

> What each part of a radiology report means, preliminary vs final reads, federal rules, what error studies show, a worked missed-nodule example and a review method.

Canonical page: https://medrecords.ai/guides/radiology-report/

---
New — **Missing Records Detection:** flags every visit, provider, and date missing from the file. [See how →](https://medrecords.ai/product/missing-records-identification/)
Record anatomy, September 2026. 10 chapters.

## Radiology report: what each part means and how to read it after the fact

For attorneys, legal nurse consultants, claims reviewers, underwriters and patients. You get the parts of a radiology report, the difference between preliminary, final and addended reads, the federal hospital and mammography rules, what research shows about misses and follow-up, a worked example with an incidental nodule and a review log.

By [Ahmed Jemaa](https://medrecords.ai/authors/ahmed-jemaa/), Co-Founder and CEO, Medrecords AI. Published 26 September 2026.

A radiology report is the radiologist's written interpretation of an imaging study such as an X-ray, CT, MRI, ultrasound or mammogram. It states the exam, the clinical question, the technique, comparison studies, the findings, and an impression that answers the question and recommends any follow-up. Federal hospital rules require the interpreting radiologist to sign it.

A radiology report is short, structured and written by a specialist, so it tends to be read as fact. In a delayed diagnosis case the question is usually different: what did the report say, where in the report did it say it, who received it, and did anyone act on it. The answers are in the report's structure, its timestamps and its distribution list.

This guide is built from the Medicare hospital Conditions of Participation for radiologic services and medical records (42 CFR 482.26 and 482.24), the CMS State Operations Manual interpretive guidance, the FDA mammography rules (21 CFR 900.12), the American College of Radiology practice parameter on communicating imaging findings (revised 2025) and published studies on radiology errors, follow-up of incidental findings, preliminary read discrepancies and malpractice claims. The worked example is hypothetical. Nothing here is legal or medical advice.

8 numbers

### The radiology report in 8 numbers

5 years
minimum a hospital must keep radiology reports and images
42 CFR 482.26(d)(2)
30 days
deadline to send a mammography report to the referring provider; 7 days if suspicious
21 CFR 900.12(c)(3)
30%
of delayed radiology diagnoses were also missed on the next imaging exam
Kim and Mansfield, AJR, 2014
251 days
average time from the first interpretation error to the correct diagnosis
Kim and Mansfield, AJR, 2014
42%
of those errors were underreading: the finding was on the image and not reported
Kim and Mansfield, AJR, 2014
29%
of incidental lung nodules needing follow-up on emergency CT scans were followed up
Blagev et al., JACR, 2014
0 of 12
nodules were followed up when the report mentioned them only in the findings section
Blagev et al., JACR, 2014
31%
of 8,401 radiologists had at least 1 malpractice claim in their career
Whang et al., Radiology, 2013
Chapter 1 Everyone

### What a radiology report is

A radiology report is the written result of an imaging study. A technologist acquires the images; a radiologist, or another authorized practitioner, reads them and dictates or types the report. The report goes to the clinician who ordered the study, and it becomes part of the medical record.

The images and the report are separate records. The images are what the scanner captured. The report is 1 physician's interpretation of them, written at 1 point in time with the history and prior studies available at that moment. When a case turns on whether something was missed, experts look at the images. Everyone else, including the treating doctors at the time, usually worked from the report.

That is why the report deserves close reading. What it says, where it says it (findings or impression), whether it recommends follow-up, when it was finalized and who it was sent to are all facts that can be established from the record without reading a single image.

#### Terms you will meet

IndicationThe clinical question or reason for the exam, usually copied from the order.
TechniqueHow the study was done: modality, body part, protocol, contrast.
ComparisonPrior studies the radiologist compared against, or a statement that none were available.
FindingsThe body of the report: what was seen, organ by organ or region by region.
ImpressionThe conclusion. Many clinicians read only this section.
Preliminary reportAn early read, often overnight by a resident or teleradiologist, that directs immediate care before the final report. Also called a wet read.
AddendumA later addition to a signed report, dated separately. The original text should remain.
Incidental findingSomething unrelated to the clinical question, such as a lung nodule on a scan ordered for a clot.

If you remember one thing
The report is 1 reading of the images at 1 moment. What it said, where, and who received it can be proved from the record alone.

Chapter 2 Everyone

### The parts of a radiology report, section by section

The mock report below is a CT scan ordered in an emergency department to rule out a blood clot in the lungs. The layout is invented. The 9 marked parts follow the structure the American College of Radiology suggests for diagnostic imaging reports.

CT angiogram, chest Hypothetical
Patient T, 58. Emergency department
Header — Exam, date and time, facility, ordering physician1
Indication — Chest pain, shortness of breath. Rule out PE2
Technique — CT pulmonary angiogram with 80 mL IV contrast3
Comparison — None available4
Findings — No filling defect. 8 mm solid nodule, right upper lobe5
Limitations — Mild respiratory motion at the lung bases6
Impression — No pulmonary embolism7
Recommendation — None stated8
Signature — Preliminary 23:55; final signed 08:10 next day9
HYPOTHETICAL
1. 1
**Header** Patient, exam, facility, date and time, and the ordering provider. The ordering provider is usually who the report is routed to, which matters when the patient's own doctor is someone else.
2. 2
**Indication** The question the study was ordered to answer. The ACR expects the report to address it.
3. 3
**Technique** Protocol, contrast and dose. A study protocoled for 1 question can be poor at answering another.
4. 4
**Comparison** Which prior studies were compared. "None available" when priors exist in the same system is a finding.
5. 5
**Findings** What was seen. Incidental findings often appear only here.
6. 6
**Limitations** Factors that reduce accuracy, such as motion or poor contrast timing. The ACR expects the report to identify them when they matter.
7. 7
**Impression** The conclusion. The section most readers rely on; a finding left out of it is easy to miss.
8. 8
**Recommendations** Follow-up imaging or other tests. The ACR expects them "when appropriate."
9. 9
**Signature and times** Dictation, preliminary, final and any addendum times, and who signed each. Federal rules require the interpreting practitioner to sign.

#### What a radiology report can and cannot tell you

| Question | Where to look | Answered by the report alone? |
| --- | --- | --- |
| What did the radiologist report? | Findings and impression | Yes |
| Was follow-up recommended? | Impression or recommendations | Yes |
| When was it final? | Signature block, audit trail | Mostly; the audit trail confirms |
| Who received it? | Distribution list, result inbox logs | Partly |
| Was an urgent finding called in? | Communication note in the report or a log | If documented |
| Was the finding visible on the images? | The images | No; that takes an expert radiologist |
| Did anyone act on it? | Clinician notes, orders, later imaging | No |

If you remember one thing
Read every section, including the findings. Note what is in the findings but missing from the impression and the recommendations.

Chapter 3 Deciding

### Preliminary, final and addended reports

Many studies produce more than 1 report. Overnight and in emergencies, a resident, a teleradiology service or an on-call radiologist gives a preliminary read so care can proceed. The final report comes later, sometimes hours later. The ACR practice parameter says a preliminary report "very likely will contain limited or incomplete information" and should be kept, labeled as preliminary, because clinical decisions may have been based on it.

| Version | What it is | What to check |
| --- | --- | --- |
| Preliminary or wet read | Early interpretation to direct immediate care | Who read it, when, who received it, and what care followed |
| Final report | The definitive interpretation, signed | Differences from the preliminary, and time between them |
| Addendum | A later addition, dated separately | Why it was added, when, and whether the original text is intact |
| Corrected report | A revised final report | What changed and who was told |
| Outside read or second opinion | Another radiologist's interpretation of the same images | Which version the treating clinicians saw |

The ACR expects a significant difference between the preliminary and final reads to be reported to the treating clinician in a way that "reasonably ensures receipt," and the communication to be documented in the final report. In a well-run department such changes are uncommon: 1 academic center found major discrepancies in 0.89% of 45,608 resident preliminary reads over a year. Uncommon is not never, and the file should show what happened when there was one.

#### Urgent and unexpected findings

The ACR describes 3 situations that call for communication beyond the routine report: findings needing immediate intervention (a collapsed lung, a misplaced tube), findings that differ from an earlier read of the same study, and significant unexpected findings that could worsen if not acted on. For these, the interpreting physician should call or otherwise reach the treating clinician, and document the date, time, method and person reached.

"Interpreting physicians should document all nonroutine communications."

If you remember one thing
Collect every version: preliminary, final, addendum. Care is often given on the preliminary, and the differences between versions need their own timeline.

Chapter 4 Everyone

### The rules behind the radiology report

Federal rules cover signing and keeping radiology reports in hospitals, and set detailed content and delivery rules for mammography. The ACR practice parameter is a professional standard, not a regulation, but it is widely cited as the benchmark for what a report should contain and how findings should be communicated.

| Source | What it requires | What it means for review |
| --- | --- | --- |
| 42 CFR 482.26(d)(1) | The radiologist or other practitioner who performs radiology services must sign reports of their interpretations | Every report needs an identifiable, signing reader |
| 42 CFR 482.26(d)(2) | Keep copies of reports and printouts, and films, scans and other image records, for at least 5 years | Images should exist for 5 years; state law may require longer |
| 42 CFR 482.24(c)(1) | Every entry legible, complete, dated, timed and authenticated | Preliminary and final times should both be recorded |
| 21 CFR 900.12(c)(1) | Mammography reports must include a final assessment category, breast density and recommendations | Check the assessment wording against the category list |
| 21 CFR 900.12(c)(2) and (3) | Lay summary to the patient and written report to the provider within 30 days; within 7 days if "Suspicious" or "Highly Suggestive of Malignancy" | Dates of the report, the letter and the provider copy |
| ACR practice parameter | Suggested report components, preliminary report handling, documented nonroutine communication | The yardstick experts use for content and communication |

[](https://www.law.cornell.edu/cfr/text/42/482.26) [](https://www.law.cornell.edu/cfr/text/42/482.26) [](https://www.law.cornell.edu/cfr/text/42/482.24) [](https://www.law.cornell.edu/cfr/text/21/900.12) [](https://www.law.cornell.edu/cfr/text/21/900.12) [](https://gravitas.acr.org/PPTS/DownloadPreviewDocument?DocId=74)
CMS guidance to surveyors adds that radiology records must include, at a minimum, the orders for the services, copies of reports and printouts, and the images. Radiology records are patient medical records, so the general medical record rules apply to them too. The mammography rules also require the report to address every clinical question the referring provider raised, "even if the assessment is negative or benign."

The ACR also says the ordering clinician shares responsibility for obtaining imaging results. In a missed follow-up case, both sides of that handoff are in the record: the report and its distribution on 1 side, the clinician's result inbox and notes on the other.

If you remember one thing
The hospital must keep the images for at least 5 years. Request them early, along with every report version and the distribution record.

Chapter 5 Deciding

### What research shows about misses and follow-up

Radiology errors are a studied subject with a large literature. A 2015 review in RadioGraphics notes that the overall rate of radiologists' errors has not changed since it was first estimated in the 1960s, and divides them into perceptual errors, where the abnormality is not seen, and cognitive errors, where it is seen but misjudged.

Kim and Mansfield reviewed 656 imaging exams with delayed diagnoses, collected at 1 center from 2002 to 2010, and classified 1,269 errors:

**The most common error types in 656 delayed radiology diagnoses** — share of 1,269 errors — Kim and Mansfield, 2014
Underreading means the finding was on the images and not reported. Satisfaction of search means the reader found 1 abnormality and stopped looking. The correct diagnosis took 251 days on average, and in 30% of cases the next exam missed it too.

The last number is the one reviewers should remember. A miss tends to repeat, partly because later readers compare against the earlier report. So a delayed diagnosis case often involves a series of reports, each of which needs its own review.

Reporting a finding does not guarantee it is acted on. Blagev and colleagues reviewed 1,000 emergency department CT scans for blood clots. 9.9% had incidental lung nodules that needed follow-up under published guidelines. Only 29% (28 of 96) were followed up. When the nodule appeared only in the findings section, the follow-up rate was 0 of 12. Even when the report explicitly recommended follow-up, it was 29% (19 of 65).

| Study | Finding | What it means for review |
| --- | --- | --- |
| Callen et al., 2012, systematic review | 1.0% to 35.7% of radiology results not followed up in ambulatory care | Follow-up failures are common enough to check for every time |
| Ruutiainen et al., 2011, 45,608 preliminary reads | 0.89% major discrepancies between resident and final reads | Rare, and worth a separate timeline when present |
| Whang et al., 2013, 8,401 radiologists | Error in diagnosis was the most common alleged cause of claims; breast cancer the most frequently missed diagnosis | Mammography and cancer follow-up files need the closest reading |

If you remember one thing
A finding in the findings section alone is often never acted on. Check the impression, the recommendations and what happened next.

Chapter 6 Building

### Worked example: an incidental nodule that was not followed up

**Hypothetical.** The patient, physicians, hospital, dates and pages are invented to show the method. It is not a real case or a customer file. Patient T, 58, a former smoker, comes to the emergency department on 03/10/2025 with chest pain. A CT angiogram rules out a blood clot. 14 months later a CT for cough shows a 2.4 cm mass in the right upper lobe, and a biopsy confirms lung cancer. The file holds the ED record (pages 1 to 8), the CT reports (pages 9 to 12), the primary care notes (pages 20 to 26) and the 2026 imaging and pathology (pages 30 to 36).

**The 2025 CT report beside what happened next** — hypothetical — Illustration
Imaging review log, 03/2025 to 05/2026, hypothetical — 9 entries
| Date | Source | Record says | Question | Page |
| --- | --- | --- | --- | --- |
| 03/10/2025 22:40 | ED order | CT angiogram chest, rule out PE | Baseline | 3 |
| 03/10 23:55 | Preliminary read | "No PE." Called to ED physician | No mention of the nodule | 8 |
| 03/11 01:30 | ED discharge note | "CT negative. Follow up with PCP" | Written before the final report | 5 |
| 03/11 08:10 | Final report | Findings: 8 mm solid RUL nodule. Impression: no PE. No recommendation | Nodule in findings only | 9 to 10 |
| 03/11 08:10 | Distribution | Routed to ED attending; PCP not listed | Who saw the final report? | 10 |
| 04/02/2025 | PCP note | "ED visit for chest pain, CT neg" | Final report not in the PCP chart | 22 |
| 04/2025 to 04/2026 | File | No chest imaging | 13-month gap |  |
| 05/04/2026 | CT chest | 2.4 cm RUL mass. "No prior studies available for comparison" | The 2025 CT was in the same system | 30 |
| 05/12/2026 | Pathology | Adenocarcinoma of the lung | The event the timeline builds toward | 34 |

Gold rows: entries that need a question. Green row: the event the timeline builds toward.
9 entries, 6 questions. The nodule was reported, but only in the findings section of a report that reached 1 physician, after the patient had gone home.

#### What each flag means

1. **A preliminary read that answered only the question asked.** "No PE" was the right answer to the order. Whether the nodule should have been in the preliminary read is a question for a radiology expert.
2. **Discharge before the final report.** The ED note calls the CT negative nearly 7 hours before the final report was signed. Any later review of the final report by the ED team would be in the ED's result inbox logs.
3. **A finding without an impression or recommendation.** The nodule is in the findings only. Whether a follow-up recommendation was expected is for the expert; the study data say findings-only nodules were rarely followed up.
4. **A distribution list without the primary care doctor.** The report went to the ordering physician. The PCP note repeats "CT neg."
5. **A later scan without comparison.** The 2026 report says no prior studies were available, though the 2025 CT is in the same system. Growth over 14 months is what a comparison would have shown.
6. **A 13-month gap.** No imaging, no mention of the nodule in any note.

#### The same case on 1 timeline

**Every dated entry that matters, ED visit to diagnosis** — hypothetical — Illustration; pages refer to the hypothetical file
1. 03/10/2025
**ED visit, CT ordered**
Rule out pulmonary embolism.

ED record p. 3
2. 03/10 23:55
**Preliminary read: no PE**
No mention of the nodule.

p. 8
3. 03/11 01:30
**Discharged home**
"CT negative."

p. 5
4. 03/11 08:10
**Final report signed**
8 mm nodule in findings; no recommendation; sent to the ED only.

pp. 9 to 10
5. 04/02/2025
**PCP visit**
"CT neg."

p. 22
6. 04/2025 to 04/2026
**No chest imaging**
13 months.

No record
7. 05/04/2026
**CT: 2.4 cm mass**
"No prior studies available."

p. 30
8. 05/12/2026
**Biopsy: lung cancer**
Adenocarcinoma.

Pathology p. 34

Read by date, the gap is not in the imaging. It is between a signed report and the physician who would have ordered the follow-up.

##### What the reports can show

From the record, without reading images

- What each version said, and where in the report.
- Whether follow-up was recommended.
- When each version was signed, and who received it.
- Whether later studies compared against earlier ones.

##### What they cannot show alone

Where you need an expert or another record

- Whether the nodule was visible and reportable (a radiologist reading the images).
- What the standard of care required in the report (an expert).
- Who opened the report (inbox and audit logs).
- What was said to the patient (testimony).

If you remember one thing
Put every report version, its signature time and its recipients on the timeline next to the clinical notes. Delays often sit in the handoff, not in the read.

Chapter 7 Building

### How to review a radiology report, step by step

The method works for a single X-ray or a series of scans over years. The goal is a list of what each report said, where, when and to whom, and what the clinicians did next.

1. **List every study.** Every imaging exam in the file, with date, modality and ordering provider.
2. **Collect every version.** Preliminary, final, addenda and outside reads for each study.
3. **Read the whole report.** Findings, limitations, impression and recommendations, in that order.
4. **Mark the gaps between sections.** Findings not carried into the impression, and recommendations that are missing or vague.
5. **Compare versions.** Differences between preliminary and final, and whether the change was communicated and documented.
6. **Check comparisons.** Whether each report compared against the prior studies that existed.
7. **Follow the report.** Distribution list, result acknowledgment, and the next clinician note that mentions it.
8. **Follow the recommendation.** Whether recommended imaging or tests were ordered, done and read.

#### Does this radiology report need a closer look?

1
**Is every abnormal finding in the findings section also in the impression?**
YesGo to 2.
NoLog each findings-only item with its page.
2
**Does the final report match the preliminary read?**
YesGo to 3.
NoLog the change, when it was made, and any documented call.
3
**Did it reach the clinician responsible for follow-up?**
YesGo to 4.
NoLog the distribution list and the first note that mentions the result.
4
**Was any recommended follow-up done?**
YesRely on the report, cited to its page, and send the images to an expert if the read itself is in question.
NoLog the recommendation and the gap to the next study.

##### 1. Radiology report checklist

The components the ACR suggests, as a checklist for reading any report.

RADIOLOGY REPORT CHECKLIST
Study: [MODALITY, BODY PART] Date/time: [ ]
Facility: [ ] Ordering provider: [ ]
Report version: [preliminary / final / addendum / corrected]

[ ] Indication stated and addressed in the impression
[ ] Technique, contrast and dose stated
[ ] Comparison studies named, or "none" (check if priors exist)
[ ] Findings: every abnormal finding listed
[ ] Limitations stated if the study was compromised
[ ] Impression answers the clinical question
[ ] Every significant finding carried into the impression
[ ] Follow-up recommendation stated where appropriate
[ ] Urgent or unexpected finding: call documented
 (date, time, method, person reached)
[ ] Dictated / preliminary / final times and signer
[ ] Distribution: who received the report
[ ] Mammography: assessment category, density,
 lay letter date, provider report date

##### 2. Imaging review log

1 line per report version, and 1 line per clinician action that follows.

IMAGING REVIEW LOG
Patient: [ID] Period: [DATES]

Date/time | Study | Version (prelim / final / addendum)
 | Findings (page) | Impression (page)
 | Recommendation | Sent to | Next action found (page)
 | Question

Example:
03/11/2025 08:10 | CT angiogram chest | Final
 | 8 mm solid RUL nodule (p. 9) | No PE (p. 10)
 | None | ED attending | PCP note "CT neg" (p. 22)
 | Findings-only nodule, PCP not on distribution

#### Before you rely on a radiology report

- **Every version is in hand** *Preliminary, final, addenda, outside reads.*
- **The images are requested** *For any read an expert will need to assess.*
- **Signature times are known** *For each version, with the signer.*
- **The distribution record is in hand** *Who received the report, and when.*
- **Prior studies are listed** *To check each comparison statement.*
- **Clinician notes after each study are in hand** *To find what was done with the result.*

0 of 6 checked.

A radiology report often leads to a biopsy; the [pathology report guide](https://medrecords.ai/guides/pathology-report/) covers what comes next. For hospital stays, the [discharge summary guide](https://medrecords.ai/guides/discharge-summary/) covers results still pending when the patient leaves, and the [EHR audit trail guide](https://medrecords.ai/guides/ehr-audit-trail-medical-malpractice/) covers how to prove who opened a result and when.

If you remember one thing
For each study, record what was said, where, when, to whom, and what happened next. The images themselves go to an expert.

Chapter 8 Deciding

### AI and radiology report review

A delayed diagnosis file can hold dozens of imaging reports over several years, each with its own versions, next to thousands of pages of clinical notes. Chapter 7's method asks for every finding, every recommendation and every follow-up action to be lined up by date. Doing that by hand is where findings-only items and missing comparisons get missed a second time.

AI medical record review can read every report in the file, pull each finding, impression and recommendation onto an AI medical chronology with a page-level citation, and put the clinician notes that follow beside them. A nodule reported in March and never mentioned again shows up as a gap you can check in seconds.

2 limits matter here. The software reads the report text, not the images; whether a finding was visible is a question for a radiologist looking at the scans. And a large language model (LLM) can hallucinate or drop a finding, which is why every line needs its page and a human in the loop checks it.

Medrecords AI works this way on the files you upload. It builds a [cited medical chronology](https://medrecords.ai/product/chronology/) across imaging, pathology and clinical notes, each entry linked to its page. You can ask "every imaging finding with a follow-up recommendation, and the next study of that area" through [cited Q&A](https://medrecords.ai/product/qa/), trace a finding across years with [condition progression tracking](https://medrecords.ai/product/condition-progression-tracking/) and [record search](https://medrecords.ai/product/medical-record-search/), and flag [reports that should exist but were not produced](https://medrecords.ai/product/missing-records-identification/). Flags are signals, not verdicts. It does not read images, decide whether a finding was missed, or retrieve records from providers. It runs under SOC 2 and HIPAA with a signed business associate agreement (see [security](https://medrecords.ai/security/) and [HIPAA](https://medrecords.ai/hipaa/) ). Self-Service bills 10 cents a deduplicated page, down to 5 cents at volume, duplicates free; Enterprise On-Prem is an annual license.

If you are comparing tools, our guide to [medical record review](https://medrecords.ai/guides/medical-record-review/) covers what to ask any medical chronology software vendor. See how imaging review fits [emergency department failure to diagnose](https://medrecords.ai/solutions/emergency-department-failure-to-diagnose-review/), [medical malpractice](https://medrecords.ai/solutions/medical-malpractice/), [expert witness](https://medrecords.ai/solutions/expert-witnesses/) and [underwriting](https://medrecords.ai/solutions/life-disability-underwriting/) work.

The offer

#### Every finding, every recommendation, and what happened next.

Book a demo on a delayed diagnosis file with years of imaging, then run your first case free on us. Every line comes back cited to its source page. You review, you revise, you sign.

[Book a demo](https://medrecords.ai/demo/) [See the cited chronology](https://medrecords.ai/product/chronology/)
Scheduling only. No records move from a public page.

If you remember one thing
AI can line up every report and what followed it, cited to the page. Whether the images showed more is a radiologist's call.

Chapter 9 Everyone

### Frequently asked questions

What are the parts of a radiology report?A header with the patient, exam and ordering provider; the indication; the technique; comparison studies; findings; limitations; the impression; recommendations; and the signature with dictation and finalization times.
What is the impression on a radiology report?The radiologist's conclusion. It should answer the clinical question and carry forward any significant finding. Many clinicians read only this section, so a finding left out of it is easy to miss.
What is the difference between a preliminary and a final radiology report?A preliminary report is an early read, often overnight, to direct immediate care. The final report is the signed, definitive interpretation. The ACR says preliminary reports should be kept because decisions may have been based on them.
What is an addendum to a radiology report?A later addition to a signed report, dated separately, such as a correction or a comparison with an outside study. The original text should remain visible.
How long must radiology images and reports be kept?Hospitals must keep reports and images for at least 5 years under federal rules. Mammography facilities have separate rules, and state law may require longer.
Who is responsible for following up on an incidental finding?The ACR says the interpreting radiologist should communicate significant unexpected findings and the ordering clinician shares responsibility for obtaining results. Who did what is established from the report, its distribution and the notes that follow.
Can a radiology report prove a finding was missed?It shows what was reported. Whether something was visible on the images and should have been reported takes a radiologist reviewing the images themselves.
Can AI read radiology reports for a legal case?AI can extract every finding, impression and recommendation from the report text and put them on a timeline with page citations. It does not replace a radiologist reading the images.
Is it HIPAA compliant to upload imaging reports to an AI tool?It can be, with a vendor that signs a business associate agreement, holds a SOC 2 report and commits in writing not to train on your data. A consumer chatbot without a BAA is a different matter.

Chapter 10 Everyone

### Sources and method

Regulations, the CMS State Operations Manual (Appendix A, revision 248) and the ACR practice parameter were read in full text from primary sources in September 2026, and quoted text is verbatim. Study figures come from the published abstracts. The mock report, patient, physicians, dates and pages in chapters 2, 6 and 7 are hypothetical. Product facts come from this site's product pages. Nothing here is legal or medical advice.

- [42 CFR 482.26](https://www.law.cornell.edu/cfr/text/42/482.26): radiologic services, signed reports and 5-year retention.
- [21 CFR 900.12](https://www.law.cornell.edu/cfr/text/21/900.12): mammography report content, patient and provider communication, and record keeping.
- [CMS State Operations Manual, Appendix A](https://www.cms.gov/regulations-and-guidance/guidance/manuals/downloads/som107ap_a_hospitals.pdf): interpretive guidance at tag A-0553 (radiology records).
- [ACR Practice Parameter for Communication of Diagnostic Imaging Findings](https://gravitas.acr.org/PPTS/DownloadPreviewDocument?DocId=74), revised 2025 (Resolution 9).
- [Kim YW, Mansfield LT, "Fool me twice: delayed diagnoses in radiology with emphasis on perpetuated errors"](https://doi.org/10.2214/AJR.13.11493), AJR 202(3):465-470, 2014.
- [Blagev DP, Lloyd JF, Conner K, et al., "Follow-up of incidental pulmonary nodules and the radiology report"](https://pubmed.ncbi.nlm.nih.gov/24316231/), Journal of the American College of Radiology 11(4):378-383, 2014.
- [Ruutiainen AT, Scanlon MH, Itri JN, "Identifying benchmarks for discrepancy rates in preliminary interpretations provided by radiology trainees at an academic institution"](https://pubmed.ncbi.nlm.nih.gov/21889753/), Journal of the American College of Radiology 8(9):644-648, 2011.
- [Whang JS, Baker SR, Patel R, et al., "The causes of medical malpractice suits against radiologists in the United States"](https://doi.org/10.1148/radiol.12111119), Radiology 266(2):548-554, 2013.
- [Callen JL, Westbrook JI, Georgiou A, Li J, "Failure to follow-up test results for ambulatory patients: a systematic review"](https://pmc.ncbi.nlm.nih.gov/articles/PMC3445672/), Journal of General Internal Medicine 27(10):1334-1348, 2012.
- [Bruno MA, Walker EA, Abujudeh HH, "Understanding and confronting our mistakes: the epidemiology of error in radiology and strategies for error reduction"](https://doi.org/10.1148/rg.2015150023), RadioGraphics 35(6):1668-1676, 2015.

### More guides

- [**Reconciling a defense record production: the page-level method for medical malpractice teams** — Why the defense production never matches your pre-suit medical records, and a page-level method to reconcile…](https://medrecords.ai/guides/reconciling-defense-record-production/)
- [**SANE exam records: the sexual assault forensic exam, page by page** — What a SANE exam record contains, how consent, evidence and chain of custody are documented, and how…](https://medrecords.ai/guides/sane-exam-sexual-assault-records/)
- [**SOAP note: what goes in each part and how to read a progress note** — What goes in each part of a SOAP note, how progress notes are written and billed, a worked example, a SOAP note…](https://medrecords.ai/guides/soap-note/)
- [**TBI imaging, diffuse axonal injury and DTI: the records that decide severity** — TBI imaging in court: why a normal CT does not rule out brain injury, how SWI and DTI get challenged, and the…](https://medrecords.ai/guides/tbi-imaging-diffuse-axonal-injury-dti/)
- [**The eggshell plaintiff rule and pre-existing conditions: how prior records decide the case** — The eggshell plaintiff rule, aggravation and apportionment explained, what the defense looks for in prior…](https://medrecords.ai/guides/eggshell-plaintiff-pre-existing-conditions/)
- [**Toxicology reports in litigation: how to read clinical and postmortem results without losing the case** — How to read a toxicology report for litigation: immunoassay screens versus GC-MS confirmation, serum alcohol…](https://medrecords.ai/guides/toxicology-report-litigation/)