X-ray (computed radiography) CT MRI PET Ultrasound Whole body

Scottish Medical Imaging (SMI) Archive

Scottish Medical Imaging (SMI) Archive and SMI Research Dataset

Population-scale copy of the Scottish national PACS: 57.3 million radiology studies (2010 to August 2018) of about 4.27 million patients, linkable to NHS Scotland health records. Approved projects analyse CT, MRI, PET and report extracts inside the Scottish National Safe Haven.

Go to dataset Edit Report

The Scottish Medical Imaging (SMI) Archive is a research copy of the national Picture Archiving and Communication System of NHS Scotland. It is held in the Scottish National Safe Haven, which the University of Edinburgh (EPCC) operates for Public Health Scotland, and was built with the Health Informatics Centre at the University of Dundee. Researchers do not download it. The eDRIS team of Public Health Scotland cuts a pseudonymised extract for each approved project, optionally linked to hospital, prescribing, cancer registry, maternity and death records through the Community Health Index number.

Composition

The snapshot described in the 2024 data resource paper covers imaging from 1 January 2010 to 31 August 2018 across all 14 NHS Scotland health boards: about 57.3 million studies, 94.9 million series and 2.47 billion DICOM images in 36 DICOM modalities. CT accounts for 3.26 million studies, MRI for 1.54 million and computed radiography for 14.2 million, and 27.2 million studies are structured reports. The paper lists 2,182,123 female and 2,081,040 male patients, and its ethnicity table adds up to 4,271,698 patients, of whom 72.1% are recorded as White. 94.2% of imaged patients have linked longitudinal health records.

Acquisition

All images come from routine clinical care, so scanners, vendors and protocols vary between health boards and over time. At publication, CT, MRI, PET and structured reports (about three quarters of all studies) were curated for research use. Public Health Scotland lists 2010 to 2017 for CT, MRI, PET and structured reports as its initial offering, with computed, digital and panoramic radiography and later years to follow.

Annotations

The paper describes no curated labels; radiologists' structured reports accompany the images. Cohorts can be built from DICOM tags, linked clinical data or natural language processing of the reports. The platform can keep annotations made by one project for reuse by later projects, subject to consent to share.

Known limitations

Access requires a project application, approval by the Public Benefit and Privacy Panel and payment of service costs, and all analysis happens inside the safe haven. Only the subset needed for the approved question is released, and trained models could not yet be exported when the paper was written. The Scottish population is about 96% White. Sorting images by sequence or body part from clinical DICOM tags alone is unreliable. Studies after August 2018 were not part of the described snapshot.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 4,271,698 subjects.

Sex

Female 2,182,123 51%
Male 2,081,040 49%

Modality

images

CT 1,787,245,067 72%
MRI 480,422,586 19%
X-ray (computed radiography) 21,193,120 <1%

License and access

Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer

Access

Application

A research proposal is reviewed and approved

Access page

Scottish Medical Imaging (SMI) access conditions (PBPP approval, Scottish National Safe Haven)

No published license. A project is approved by the NHS Scotland Public Benefit and Privacy Panel, then works on a project-specific extract inside the Scottish National Safe Haven, paid on a cost recovery basis. AI development is allowed, the data cannot be exported and outputs pass disclosure control. Commercial AI products may have to be offered to NHS Scotland at a discount.

Original license text Version read: Public Health Scotland SMI pages last updated 2025-08-19, and the access conditions described in Baxter et al., Radiology: Artificial Intelligence 6(1):e220266 (2024), both read 2026-10-09 Checked 2026-10-09

What you can do

Conditional
Yes
Yes
Conditional

What you can share

No
Conditional
No

What you must do

Not stated
Share alike Not stated
Not stated
Not stated
Manuscript review Not stated
Release code Not stated
Return results Not stated
Delete after use Not stated

Limits

Yes
Conditional

Citation

Baxter R, Nind T, Sutherland J, et al. The Scottish Medical Imaging Archive: 57.3 Million Radiology Studies Linked to Their Medical Records. Radiology: Artificial Intelligence 6(1):e220266 (2024). doi:10.1148/ryai.220266

@article{baxter2024,
  title={The Scottish Medical Imaging Archive: 57.3 Million Radiology Studies Linked to Their Medical Records},
  author={Baxter, Rob and Nind, Thomas and Sutherland, James and others},
  journal={Radiology: Artificial Intelligence},
  volume={6},
  number={1},
  pages={e220266},
  year={2024},
  doi={10.1148/ryai.220266}
}

All numbers

Every number on this page, as stored in stats.csv, with its source.

Measure	Breakdown	Value	Source
Subjects	total sum of the nine disjoint ethnicity rows of Table 2, whose percentages add up to 100%; the paper gives no single patient total	4,271,698	baxter2024 Table 2
Subjects	sex=female	2,182,123	baxter2024 p. 2 (Data Resource Collection)
Subjects	sex=male	2,081,040	baxter2024 p. 2 (Data Resource Collection)
Studies	total 57.3 million studies, all modalities, January 2010 to August 2018	~57,300,000	baxter2024 p. 1 (Data Resource Collection)
Studies	sex=female unique study identifiers of female patients	13,167,604	baxter2024 p. 2 (Data Resource Collection)
Studies	sex=male unique study identifiers of male patients	11,265,313	baxter2024 p. 2 (Data Resource Collection)
Studies	modality=CT	3,261,004	baxter2024 Table 1
Studies	modality=MR	1,539,189	baxter2024 Table 1
Studies	modality=CR computed radiography; not yet curated for research access at publication	14,182,921	baxter2024 Table 1
Scans	total 94.9 million DICOM series across all modalities, including structured reports	~94,900,000	baxter2024 p. 1 (Data Resource Collection)
Scans	modality=CT DICOM series	12,909,599	baxter2024 Table 1
Scans	modality=MR DICOM series	15,878,418	baxter2024 Table 1
Images	total 2.47 billion DICOM images, all modalities	~2,470,000,000	baxter2024 p. 1 (Data Resource Collection)
Images	modality=CT DICOM images (slices)	1,787,245,067	baxter2024 Table 1
Images	modality=MR DICOM images (slices)	480,422,586	baxter2024 Table 1
Images	modality=CR DICOM images	21,193,120	baxter2024 Table 1

Sources

The keys used in the table above.

baxter2024 Baxter R et al. 2024, The Scottish Medical Imaging Archive, Radiology Artificial Intelligence (CC BY 4.0, version of record in the University of Dundee repository) paper
phs-smi-catalogue Public Health Scotland, Scottish Medical Imaging (SMI), SMI catalogue page website
epcc-smi-service EPCC, Scottish Medical Imaging Service announcement website

Overview