Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0638
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-04T06:07:28.683534Z
status: published
lens: trace
sector: health
headline: Promises and limitations of deep learning for predicting knee osteoarthritis progression from medical imaging: A systematic review
dek: To systematically evaluate the performance, methodological quality, and translational barriers of deep learning (DL) models for predicting knee osteoarthritis (KOA) progression from medical imaging. Following PRISMA guidelines, we searched PubMed, Scopus, and Web of Science (inception to June 2026) for peer-reviewed studies applying DL to predict KOA progression from medical imaging. Two reviewers independently screened studies, extracted data, and assessed risk of bias using PROBAST-AI. The primary outcome was…
gain_title: Deep learning models demonstrated proof-of-concept ability to predict knee osteoarthritis progression from medical imaging, with internal median AUCs up to 0.87 for surgical endpoints.
problem_title: Deep learning models for knee osteoarthritis progression showed limited generalizability, with performance degradation on external validation and heavy reliance on a single training dataset without rigorous multi-site validation.
trace_subject: deep learning models predicting knee osteoarthritis progression from medical imaging
gain_reading: Deep learning models demonstrated proof-of-concept ability to predict knee osteoarthritis progression from medical imaging, with internal median AUCs up to 0.87 for surgical endpoints.
gain_evidence: DL models demonstrate proof-of-concept for predicting KOA progression | Surgery prediction models showed the highest descriptive median AUC (0.87) | MRI-based models showed higher internal median AUCs (0.85) than radiograph-based models (0.81)
problem_reading: Deep learning models for knee osteoarthritis progression showed limited generalizability, with performance degradation on external validation and heavy reliance on a single training dataset without rigorous multi-site validation.
problem_evidence: Externally validated models showed performance degradation (median AUC: 0.75 vs. 0.81 internal) | only 27% performed external validation | Nearly all studies (97%) used the Osteoarthritis Initiative dataset for training
quick_read: This PRISMA systematic review evaluated 33 peer-reviewed studies (2019-2026) comprising 50 deep learning models that predict knee osteoarthritis progression from medical imaging. It extracted AUC as primary outcome, categorized nine different progression definitions, and assessed bias with PROBAST-AI, finding median internal AUCs of 0.87 for surgery, 0.78 for structural, and 0.79 for symptomatic endpoints.

The findings matter because they show proof-of-concept predictive signal but highlight that clinical deployment is premature. As of the August 2026 publication date, nearly all models were trained on the Osteoarthritis Initiative dataset, few were externally validated, and those that were degraded in performance, leaving uncertainty about generalizability, standardized outcome definitions, and true clinical utility.
limitation: Review found substantial heterogeneity preventing meta-analysis, heavy reliance on single dataset, limited external validation, and high risk of bias in analysis domain, indicating models not ready for clinical deployment.
tag: Automated dual reading
key_points: Systematic review of 33 studies (2019-2026) encompassing 50 predictive models with median sample size 4298 knees. | Nine progression definitions identified: structural deterioration (n=14), symptomatic worsening (n=2), surgical endpoints (n=8), combined outcomes (n=9). | 97% of studies used Osteoarthritis Initiative dataset for training; only 27% performed external validation. | Externally validated models degraded to median AUC 0.75 versus 0.81 internal; 76% had unclear or high risk of bias in analysis domain per PROBAST-AI.
rundown: The review searched PubMed, Scopus, and Web of Science to June 2026 and included 33 studies with 50 models. Sample sizes ranged from 340 to 52,981 knees. Progression was defined variably across studies, and combining imaging modalities did not consistently improve predictions, while studies leveraging longitudinal trajectories reported higher internal performance.

Risk of bias assessment using PROBAST-AI found low risk in participant selection and outcome assessment for most studies, but 76% had unclear or high risk in the analysis domain. Authors concluded that overcoming translational barriers requires standardised progression definitions integrating structural and symptomatic outcomes, rigorous multi-site validation, and models that effectively leverage multimodal and longitudinal data.
sources:
- peer_reviewed | Knee Surgery, Sports Traumatology, Arthroscopy | https://doi.org/10.1002/ksa.70555 | 2026-08-03
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

04bb6362d63e8e940eb3cedce0e89c0d809f5ffd5fb3d0db679cce6e5230a76b

previous

0000000000000000000000000000000000000000000000000000000000000000