Integrating Radiomics and Neural Networks for Knee Osteoarthritis Incidence Prediction

Li, S; Cao, P; Li, J; Chen, T; Luo, P; Ruan, G; Zhang, Y; Wang, X; Han, W; Zhu, Z; Dang, Q; Wang, Q; Zhang, M; Bai, Q; Chai, Z; Yang, H; Chen, H; Tang, M; Akbar, A; Tack, A; Hunter, DJ; Ding, C

doi:10.1002/art.42915

Osteoarthritis

Integrating Radiomics and Neural Networks for Knee Osteoarthritis Incidence Prediction

Arthritis RheumatolPublished 8 Jul 2024Diagnostic accuracy study

Read the original article ↗PubMed

Cite

Authors' conclusion

JS-RM integrating the features of the meniscus and cartilage showed improved predictive values in radiographic KOA incidence.

Study context

Population: Osteoarthritis Initiative cohort: knees without radiographic knee osteoarthritis (KOA) at baseline but at high risk; patients' knees developed radiographic KOA, control knees did not over four years549 development, 137 test knees

Comparison: Resident surgeons' magnetic resonance imaging (MRI) reading without vs with JS-RM aid
Outcome: Prediction of radiographic KOA incidence (AUC, sensitivity, specificity)
Follow-up: Four years

Methodological detailsradiomic model with neural networks, random split into development and test cohorts (8:2), reader experiment

Radiomic model with neural networks
Random split into development and test cohorts (8:2)
Reader experiment

Performance

As reported in the abstract

Prediction of radiographic KOA incidence, test cohort

Test or model	AUC	Sensitivity	Specificity
JS-RM	0.931 (95% CI 0.876-0.963)	84.4% (95% CI 83.9%-84.9%)	85.6% (95% CI 85.2%-86.0%)

JS-RM

AUC0.931 (95% CI 0.876-0.963)

Sensitivity84.4% (95% CI 83.9%-84.9%)

Specificity85.6% (95% CI 85.2%-86.0%)

Further findings

Outcome	Without JS-RM	P value
Resident surgeons' mean specificity and sensitivity predicting radiographic KOA incidence (reader experiment)With JS-RM vs Without JS-RM · specificity and sensitivity		P < 0.001
Mean specificity	0.474 (95% CI 0.333-0.614)
Mean sensitivity	0.586 (95% CI 0.429-0.743)

Resident surgeons' mean specificity and sensitivity predicting radiographic KOA incidence (reader experiment): With JS-RM mean specificity 0.874 (95% CI 0.847-0.901) · mean sensitivity 0.812 (95% CI 0.742-0.881)

Original publication

Integrating Radiomics and Neural Networks for Knee Osteoarthritis Incidence Prediction. Arthritis Rheumatol. Published 8 July 2024. doi:10.1002/art.42915

Read the original article ↗

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