Table 3.

Detailed comparison of the developed models for knees identified with Kellgren-Lawrence grade 0 or 1, which is considered as absence of osteoarthritis.

Model #	Model	AUC	AP
2	Age, Sex, BMI, Injury, Surgery, WOMAC, KL-grade (LR)	0.73 (0.70–0.75)	0.52 (0.49–0.55)
4	Age, Sex, BMI, Injury, Surgery, WOMAC, KL-grade (GBM)	0.75 (0.72–0.77)	0.54 (0.51–0.58)
5	CNN	0.78 (0.76–0.80)	0.58 (0.55–0.61)
6	CNN + Age, Sex, BMI, Injury, Surgery, WOMAC (GBM-based fusion)	0.78 (0.76–0.80)	0.58 (0.55–0.62)
7	CNN + Age, Sex, BMI, Injury, Surgery, WOMAC, KL-grade (GBM-based fusion)	0.80 (0.78–0.82)	0.62 (0.58–0.65)

The testing was done on the Multicenter Osteoarthritis Study dataset. 95% confidence intervals are reported in parentheses for each of the reported metric.

KL-grade – Kellgren-Lawrence grade. CNN – Deep Convolutional Neural Network. BMI – Body-Mass Index. WOMAC – Western Ontario and McMaster Universities Arthritis Index. AUC – Area Under the Receiver Operating Characteristic Curve. AP – Average Precision. LR – Logistic regression. GBM – Gradient Boosting Machine.