Table 3.
Diagnostic and predictive models built using protein spectra.
| AI methods used | Authors [ref.] | Spectrometry or spectroscopy method | Stage of endometriosis | Type of endometriosis | Sample size | Inputs used | Method accuracy |
|---|---|---|---|---|---|---|---|
| Support Vector Machines | Parlatan et al.37 | Raman Spectroscopy | All four stages of endometriosisa | Not specified | 94 serum samples (49 endometriosis, 45 controls) | 790–1729 cm−1 spectral interval |
SE = 87.5% SP = 100% |
| k-nearest neighbor (weighted) | Parlatan et al.37 | Raman Spectroscopy | All four stages of endometriosisa | Not specified | 94 serum samples (49 endometriosis, 45 controls) | 790–1729 cm−1 spectral interval |
SE 100% SP = 100% |
| Partial least squares discriminant analysis (PLSDA) | Braga et al.36 | Mass Spectrometry | Stage 3 and 4 | Not specified | 100 patients (50 endometriosis, 50 controls) | Positive ionization m/z = 758.7234, 786.7585, 758.7155, 782.7239, 369.4541; negative ionization m/z = 279.3316, 215.1182, 255.3261, 281.3487, 283.36375 |
SE = NR SP = NR |
| Dutta et al.42 | 1H-NMR Spectroscopy | Stage 1 and 2 | Not specified | 45 patients (22 endometriosis, 23 controls) | TSP, lipoproteins (LDL and VLDL), unsaturated lipid, creatinine, L-Arginine, glucoerophosphatidylcholine, D-glucose, ornithine, citrate, L-lysine, tyrosine, L-histidine, L-phenylalanine, formate, choline, L-threonine, acetate, L-glutamine, succinate, acetone, adipic acid, L-isoleucine, alanine, L-aspartate, 3-hydroxybutyric acid, propylene glycol, valine, leucine, creatine, pyruvate, lactate, 2-hydroxybutyrate |
SE = 81.8% SP = 91.3% |
|
| Quadratic discriminant analysis | Ghazi et al.38 | Nuclear magnetic resonance spectroscopy | Stage 2 and 3 | Not specified | 45 patients (31 endometriosis, 15 controls) | Chemical shift for all spectra between 0 to 5.5ppm |
SE = NR SP = NR |
| Genetic algorithm | Wang et al.39 | Liquid chromatography tandem mass spectrometry | All four stages of endometriosisa | Not specified | 122 patients (60 endometriosis, 62 without endometriosis) | m/z = 1433.9, 1599.4, 2085.6, 6798, 3217.2 |
SE = 90.9% SP = 92.9% |
| Wolfler et al.43 | Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry | Not specified | Not specified | 91 symptomatic patients | Mass peaks between 2000 and 20000 Da |
SE = 55.6% SP = 64.9% |
|
| Decision tree algorithm | Wang et al.39 | Liquid chromatography tandem mass spectrometry | All four stages of endometriosisa | Not specified | 122 patients (60 endometriosis, 62 without endometriosis) | 36 differentially expressed peptide spectra |
SE = 90% SP = 80.6% |
| Wolfler et al.43 | Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry | Not specified | Not specified | 91 symptomatic patients | Mass peaks between 2000 and 20000 Da |
SE = 92.7% SP = 62.8% |
|
| Quick classifier algorithm | Wang et al.39 | Liquid chromatography tandem mass spectrometry | All four stages of endometriosisa | Not specified | 122 patients (60 endometriosis, 62 without endometriosis) | 36 differentially expressed peptide spectra |
SE = 73.3% SP = 77.4% |
| Least squares support vector machines | Fassbender et al.40 | Matrix-assisted laser desorption ionization time-of-flight mass spectrometry | Stage 1/2, stage 3/4 | U/S negative endometriosis | 254 plasma samples (165 endometriosis, 89 without endometriosis) | Minimal to mild endometriosis m/z = 4898, 5715, 8328, 9926, 14.698; moderate to severe endometriosis m/z = 3192, 4519, 2189, 4373, 7457; ultrasonography-negative endometriosis m/z = 2.058, 2456, 3.883, 14.694, 42.065 |
Minimal to mild endometriosis: SE = 75% SP = 86% Moderate to severe endometriosis: SE = 98% SP = 81% Ultrasonography-negative endometriosis: SE = 88% SP = 84% |
| Fassbender et al.41 | Proteomic surface-enhanced laser desorption ionization time-of-flight mass spectrometry | All four stages of endometriosisa | Not specified | 49 endometrial biopsies (31 endometriosis, 18 without endometriosis) | m/z = 2072, 2973, 3623, 3680, 21113 |
SE = 91% SP = 80% |
|
| Artificial neural networks | Ghazi et al.38 | Nuclear magnetic resonance spectroscopy | Stage 2 and 3 | Not specified | 45 patients (31 endometriosis, 15 controls) | Chemical shift for all spectra between 0 and 5.5ppm |
SE = 50% SP = 17% |
| Wang et al.26 | Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry | All four stages of endometriosisa | Not specified | 39 patients (26 endometriosis, 13 controls) | m/z = 6898, 5891, 5385, 6448, 5425 |
SE = 91.7% SP = 90.9% |
|
| Wang et al.44 | Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry | All four stages of endometriosisa | Not specified | 66 serum samples (36 endometriosis, 30 controls) | m/z = 8142, 5640, 5847, 8940, 3269 |
SE = 91.7% SP = 90% |
NR not reported, m/z mass-to-charge ratio, ppm parts per million, Da Dalton, TSP thrombospondin, VLDL very-low-density lipoprotein, LDL low-density lipoprotein, 1H-NMR hydrogen-1 nuclear magnetic resonance, U/S ultrasound, SE sensitivity, SP specificity.
aMinimal, mild, moderate and severe stages of endometriosis were included.