Musculoskeletal (MSK)AI / InformaticsResearch

LightGBM model predicts post-arthroplasty hypoproteinemia with external AUC 0.855

Radiology AI literature (PubMed)Jun 12

Multicenter retrospective: LightGBM model predicts post-TJA hypoproteinemia (external AUC 0.855). Age, erythrocyte sedimentation rate, surgery time, serum calcium, and anesthesia method were key predictors. Not yet prospectively validated.

  • Multicenter retrospective study in 1,383 patients (training 585, internal validation 250, external validation 548).
  • Model achieved training AUC 1.000, internal validation AUC 0.968, external validation AUC 0.855.
  • Retrospective design; requires prospective validation before clinical use.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.