Osteoarthritis Initiative

The Osteoarthritis Initiative is an NIH-led, public-private observational study that followed people with or at risk of knee osteoarthritis for more than a decade. Its clinical data and images have been shared since 2004 and now live in the NIMH Data Archive. The imaging is widely used to study cartilage loss and disease progression, to grade radiographic severity, and to train knee MRI segmentation and X-ray grading models.

Composition

The cohort has 4,796 participants, 58% of them women, aged 45 to 79 at enrolment. It is made of three groups: a progression cohort with symptomatic tibiofemoral osteoarthritis (1,389), an incidence cohort without symptomatic disease but with risk factors (3,285), and a small control cohort with neither symptoms, radiographic disease nor risk factors (122). In-clinic visits with imaging took place at baseline and at 12, 24, 36, 48, 72 and 96 months; subsets of the progression cohort had extra knee MRI at 18 or 30 months.

Acquisition

Five US centres ran the study. Each knee was scanned separately at 3T with a sagittal 3D DESS sequence, coronal and sagittal intermediate-weighted turbo spin echo, and, for one knee, a coronal 3D FLASH and a multi-echo spin-echo T2 map. Bilateral fixed-flexion knee radiographs were taken at each imaging visit; pelvis, hand and full-limb radiographs and thigh MRI were added at selected visits.

Annotations

Central groups produced semi-quantitative and quantitative image assessments: Kellgren-Lawrence and OARSI grades and joint space width on radiographs, WORMS, BLOKS or MOAKS scores on MRI, cartilage volume and thickness measurements, and limb alignment angles. These cover subsets of participants and time points.

Known limitations

• Image counts per visit and sequence are not published in an open document, so this page lists only cohort totals.
• Image assessments exist only for selected subsets and visits.
• Access needs an NDA account and a data access request; the data use terms forbid passing on subject-level data.