Skip to content

Specialized data (특화 데이터)

Specialized data are research datasets built by the medical-data-centered hospitals (의료데이터 중심병원). They are organized by disease and by hospital, and they include material the public library never has, such as images, waveforms, and curated clinical tables. The portal home page, on 29 October 2026, counted 67 of them.

Sources

The definition and the application steps follow the portal's specialized-data page. The hospital program that produces the datasets is described on the portal's medical-data-centered hospital page. The count of 67 follows the portal home page on 29 October 2026.

What they cover

The portal describes the program as building high-quality Korean datasets for research and for training models, in cancer, circulatory disease, and respiratory disease including COVID-19, plus other disease-, age-, and patient-group sets. A consortium page lists, as examples, cancer imaging and endoscopy, brain MRI, intensive-care waveforms, chest X-ray, and pathology slides. Each dataset has its own hospital, years, patient count, and table count. This site does not yet reproduce that catalog.

These datasets are not the Cancer Public Library and not the national clinical-library schema. A lung-cancer image set from one consortium is that consortium's files. It does not contain the national registry, unless the dataset's own description says it was linked.

How access works

The portal lists six steps, and they differ from the public-library timetable.

  1. Read the dataset overview, the items, and the application rules on the portal.
  2. Apply under the providing institution's procedure.
  3. The providing institution reviews release. Moving the data to a safe-use center at a different institution is a further review.
  4. Agree the place where the data will be analyzed.
  5. Analyze, and request export of the results.
  6. The institution that runs the analysis room reviews the export.

A study that only needs the national registry should use the Cancer Public Library. A study that needs images or a hospital's curated cohort uses this menu, and has to read that dataset's own page.