Solutions / Custom datasets

Custom medical imaging datasets for AI

MedCorpora scopes custom medical imaging datasets around a buyer's anatomy, pathology, modality, acquisition, population, clinical ground truth and permitted use. Matching hospital inventory and authority are verified before an approved, de-identified cohort is curated and released.

Published 30 July 2026 · Reviewed 12 August 2026 · MedCorpora
ModalitiesMRI · CT · X-ray · ultrasound
DesignBuyer-defined cohort
EvidenceReports · labels · outcomes
FormatDICOM · traceable derivatives
01

What the programme covers

A custom programme is appropriate when a public benchmark or fixed catalogue cannot represent the target clinical setting. The cohort can be narrowed by modality, body region, disease, scanner, protocol, demographic distribution, label source and intended development or validation role.

02

What a useful specification includes

A defensible request defines the clinical task, source evidence and acceptance criteria before patient-level data moves. The exact fields and thresholds depend on the intended model claim.

  • Target anatomy, pathology and model task
  • Required modality, sequences, views, protocols or reconstruction
  • Positive, negative, difficult-case and demographic composition
  • Clinical truth, annotation and reviewer requirements
  • Development, evaluation or external-validation split
  • Permitted commercial use, geography and retention
03

Quality and validation controls

Imaging quality is assessed at examination and series level rather than by raw file count. Derived formats remain mapped to source DICOM and cohort exclusions are documented.

  • Study and series completeness
  • Scanner, protocol and acquisition distributions
  • Duplicate and longitudinal-patient controls
  • Label provenance and adjudication state
  • Metadata, pixel and facial-anatomy privacy checks
04

Availability, rights and delivery

A pilot sample can be assessed against written criteria before the final cohort is compiled, licensed and delivered through the agreed secure channel.

Public pages describe a sourcing and engineering capability, not guaranteed ready inventory. Each release remains subject to verified programme inventory, programme-specific authorization, privacy review, technical acceptance and buyer licence terms.

05

Questions, answered directly.

Can the dataset combine several hospitals?

Potentially. Multi-site composition depends on matching inventory, permissions and a consistent release specification at each participating source.

Are annotations automatically included?

No. Reports or diagnoses may be available, while task-specific masks, boxes or measurements require separate clinical annotation scope.

Can data be delivered as NIfTI?

Yes for appropriate imaging programmes, with a traceable mapping to source DICOM and recorded conversion rules.

Is a minimum order required?

Volume and commercial structure are established after feasibility and quality work are understood.

Institutional engagement

Define the cohort.