Osteoarthritis

Predicting knee osteoarthritis radiographic progression: A multimodal diffusion approach

Osteoarthritis CartilagePublished Model development and validation study

Authors' conclusion

KOA-Diff synthesizes future radiographs reflecting accurate structural progression, providing an interpretable tool for KOA prediction.

Study context

Population
People with knee OA: Osteoarthritis Initiative (OAI) dataset and a private external cohort (PUTH-KOA)external cohort n=80
Comparison
OAI vs PUTH-KOA
  • PUTH-KOAn = 80
Outcome
Accuracy of synthesized future knee radiographs: joint space width error (JSWE) and binary KL-grade accuracy (early [KL ≤2] vs advanced [KL ≥3])
Methodological detailslatent diffusion model with multimodal fusion network, internal validation (OAI), external validation (PUTH-KOA), blinded expert evaluation
  • Latent diffusion model with multimodal fusion network
  • Internal validation (OAI)
  • External validation (PUTH-KOA)
  • Blinded expert evaluation

Performance

As reported in the abstract

JSWE and binary KL-grade accuracy of synthesized radiographs · over 96 months (OAI)
Test or modelJoint space width error (JSWE)Binary Kellgren-Lawrence (KL)-grade accuracy
OAI0.71%-1.21%84.0%-91.8%
PUTH-KOA2.06%90.0%

Joint space width error (JSWE)

OAI0.71%-1.21%
PUTH-KOA2.06%

Binary Kellgren-Lawrence (KL)-grade accuracy

OAI84.0%-91.8%
PUTH-KOA90.0%

Further findings

OutcomeOAIPUTH-KOA
Blinded expert rating of synthesized feature accuracy (5-point scale)
Joint space narrowing4.1624.050
Osteophyte formation3.8953.925
Subchondral bone sclerosis3.9433.775

Clinicians' accuracy identifying high-risk patients (KL grade increase ≥2 levels over 8 years), without vs with model assistance

Accuracy without model assistance
58.3%
Accuracy with model assistance
77.5%

Original publication

Predicting knee osteoarthritis radiographic progression: A multimodal diffusion approach. Osteoarthritis Cartilage. Published 31 August 2026. doi:10.1016/j.joca.2026.08.013

Composed by AI from the published abstract and checked against it. How Briefs are made →

Get them every week

New studies on this topic in My Week and in your Monday e-mail.

We e-mail you a sign-in link and a code. No password, no spam.