About GAMA-RADS

What this is

GAMA-RADS — Imaging Repository & Advanced Processing — exists because two things were happening separately and badly. Teaching cases lived in personal folders and presentation decks, findable only by the person who collected them. Research imaging was re-processed from scratch for every study, with pipeline settings recorded in whatever the analyst remembered to write down.

The platform puts both on one footing: a repository where cases carry structured findings and are searchable by modality, region and diagnosis, and a pipeline that records exactly what was run, in what order, with which parameters, so a cohort can be rebuilt years later and produce the same numbers.

The department

Departemen Radiologi sits within Fakultas Kedokteran, Kesehatan Masyarakat dan Keperawatan, Universitas Gadjah Mada, in Yogyakarta. It trains radiology residents, runs diagnostic imaging services, and conducts imaging research across neuroradiology, thoracic, musculoskeletal, abdominal and paediatric subspecialties.

The department owns this platform: it approves what enters the repository, decides who gets access, and answers for how the data is handled. That ownership is the point — a repository of patient imaging governed by nobody in particular is not one anybody should contribute to.

What is in the repository

Cases are contributed by departmental staff and residents under ethics approval, de-identified at ingest, and reviewed before they become visible. Each carries structured findings rather than free text, so a search for a finding returns cases rather than a keyword match.

• Teaching sets — curated around a topic or an examination syllabus, annotated for learners.
• Board preparation — cases selected for their discriminating features, with the reasoning made explicit.
• Research cohorts — assembled to a protocol's inclusion criteria and versioned, so a published result can be traced to an exact set of studies.

How the pipeline works

Five stages, each recording its inputs, parameters and outputs: ingest from PACS or disk, de-identification with verification, preprocessing (resample, normalise, register), model-assisted segmentation reviewed by an expert, and export as a versioned cohort. Any output can be traced back through those stages to the studies it came from.

Model assistance is exactly that — assistance. Every segmentation that reaches the repository has been reviewed and corrected by someone qualified to do it.

Governance

Access is granted per account after institutional review, at the narrowest tier that answers the question being asked. Contributions require ethics approval. Releases beyond the platform require a data use agreement. Authorisation is enforced on every request by the database itself; a per-case access log is a gap we name in the privacy notice rather than claim as done.

The detail is on the data governance page, including the de-identification profile and the access tiers.

Getting access

Request an account with your institutional email, stating your role and what you intend to use the repository for. Residents and students are sponsored by a supervisor. Researchers supply their protocol approval reference.

Applications are reviewed by the department, normally within five working days. Read the terms of use first — particularly the parts about clinical use and re-identification, which are the two conditions that matter most.

Status of this build

The parts that are real: the design system, the component library, the accessibility work described on the accessibility page, and the imaging shown in the case viewer, which is de-identified departmental imaging used as visual reference.