Table 4 Comparison of classification results with literature.
From: A machine learning approach for non-invasive PCOS diagnosis from ultrasound and clinical features
Reference | Dataset | Best method | Accuracy | AUC | Features |
|---|---|---|---|---|---|
Zad et al. | 30,601 patients | Gradient Boost | NA | 85% | Age, Gravity, HDL, Race, Obesity, bHCG, BMI, Hypertension |
Elmannani et al. | 541 patients | Stacking ML | 96% | 94% | Follicle numbers on both ovaries, Cycle, Weight, Age, Hip, Vit D3, PRL, CL |
Tiwari et al. | 541 patients | RF | 92% | NA | Hair growth, Follicle numbers on both ovaries, Pimples, Waist, |
Rahman et al. | 541 patients | RF, AdaBoost | 94% | 89% | All features in the dataset |
Khanna et al. | 541 patients | Multi Stacking ML | 98% | NA | Follicle numbers on both ovaries, Skin darkening, Fast food, Hair growth’ Cycle, FSH/LH, Cycle length, Weight gain, AMH, PRL, Pimples, BP Systolic, Waist, Age |
Abu Abda et al. | 541 patients | Linear SVM | 91.60% | NA | Clinical, metabolic imaging, hormonal & biochemical |
Ahmetasevic et al. | 1,000 patients | ANN | 96% | NA | All features in the dataset |
Current study | 541 patients | XGBoost | 96% | 99.5% | Skin darkening, Hair growth, Weight gain, Menstrual cycle irregularities, Follicle count on both ovaries, and AMH |
