Umów wizytę

Tomasz Poboży

AI for femoral head osteonecrosis: high diagnostic accuracy, limited external validation

Autor:

Tomasz Poboży

Polish version: AI w diagnostyce martwicy głowy kości udowej — wysoka trafność, ale wciąż mało walidacji zewnętrznej

Artificial intelligence algorithms have achieved high accuracy for detecting femoral head osteonecrosis on imaging. A meta-analysis published in the Journal of Medical Internet Research on 20 August 2026 reported pooled sensitivity of 0.91 and specificity of 0.95. These findings do not establish that an algorithm can replace radiologists or treating clinicians.

Why this study matters

Early osteonecrosis may be invisible on radiographs, while MRI can identify disease before articular collapse. The clinical question is whether algorithms improve diagnostic consistency, help detect early lesions and remain reliable beyond the centre where they were developed. Strong performance on local data may not transfer to another hospital, scanner or patient population.

Review methods

The systematic review and meta-analysis included 12 studies involving 16,189 hips. It pooled sensitivity, specificity and area under the ROC curve (AUC) for imaging-based algorithms and examined sources of between-study variation. Subgroups included MRI versus radiographs and internal versus external validation. Testing on independent external data is a more demanding assessment of generalisability.

Main results

  • Pooled sensitivity was 0.91 (95% CI 0.87–0.95) and specificity 0.95 (95% CI 0.93–0.96).
  • AUC was 0.97 (95% CI 0.95–0.98), indicating strong discrimination.
  • For early disease, sensitivity was 0.93, specificity 0.94 and AUC 0.98.
  • MRI models had the strongest diagnostic performance, with a substantially higher diagnostic odds ratio than radiograph-based models.
  • Externally validated studies had less striking results than those relying mainly on internal validation.

Heterogeneity was substantial, and imaging modality explained a large share of the variation. There is no single universal performance figure independent of input data and model design.

Clinical interpretation for an orthopaedic surgeon

Algorithms may support imaging interpretation, particularly MRI, where high sensitivity could help avoid missing early disease and high specificity could limit false positives. Research accuracy does not establish readiness for autonomous clinical decisions. Diagnosis still requires symptoms, risk factors and disease staging to be considered together with imaging.

Related background on pobozy.org includes femoral head osteonecrosis after COVID-19 (Polish) and steroid- and alcohol-related osteonecrosis (Polish).

Limitations

  • Only 12 studies were included, most retrospective.
  • Populations, imaging, model architectures and reference standards varied substantially.
  • External validation remained relatively uncommon.
  • Diagnostic accuracy does not establish effects on treatment decisions, time to diagnosis, costs, missed lesions or clinical outcomes.
  • The review does not demonstrate equivalence to experienced radiologists or safety as a standalone diagnostic method.

Implications for practice

The findings support further development of imaging assistance, followed by prospective multicentre testing on independent data and assessment of its actual effect on diagnostic care. They do not justify omitting clinical examination, avoiding MRI when early osteonecrosis is reasonably suspected, replacing clinicians or assuming equal performance across all available models. The most credible role remains support for clinical judgement.

References

  • Poboży T, Poboży K, Domańska-Poboża J, Konarski W. Impact of COVID-19 on the Development of Femoral Head Avascular Necrosis: A Systematic Review. Medical Sciences. 2026;14(3):372. DOI: 10.3390/medsci14030372.
  • Konarski W, Poboży T, et al. Osteonecrosis Related to Steroid and Alcohol Use—An Update on Pathogenesis. Healthcare. 2023;11(13):1846. DOI: 10.3390/healthcare11131846.
  • Lu F, Wang L, Zhang W, Ma Y, Tian J, Hu Y. Diagnostic Accuracy of Medical Imaging-Based Artificial Intelligence for Osteonecrosis of the Femoral Head: Systematic Review and Meta-Analysis. Journal of Medical Internet Research. 2026. DOI: 10.2196/95648. PMID 42623169.

Related resources

For Doctors and Physiotherapists · Research Updates, Literature and Evidence-Based Medicine

Content prepared by: Tomasz Poboży - orthopaedic surgeon.

Więcej wpisów