Table 1.
Summary of studies investigating ML methods applied for the diagnosis of rare diseases.
| Disease | Methods | Data Type | Sample Size | Model Performance | References |
|---|---|---|---|---|---|
| Huntington disease | ANN | footstep pressure data | 180 | ACC: 89% | [42] |
| SVM | gait data | 15 (post stroke patients), 17 (HD patients), 10 (Controls) | ACC: 90.5% | [43] | |
| SVM | eye tracking data | 22 (Controls), 14 (pre-HD patients), 14 (HD patients) | ACC: 73.47%, TPR: 74.31%, TNR: 72.64% pre-HD vs. Controls ACC: 81.84%, TPR: 76.19%, TNR: 87.48% HD vs. Controls ACC: 83.54%, TPR: 92.62%, TNR: 74.45 % pre-HD vs. HD |
[44] | |
| BM | quantitative electroencephalography data | 26 (preHD patients), 25 (Controls) | TNR: 83%, TPR: 83%, ACC: 83%. | [45] | |
| Multiple system atrophy | SVM | MRI data | 30 (MSA patients), 62 (PD patients), 59 (Controls) | ACC: 77.17% | [46] |
| SVM | MRI data | 31 MSA patients, 65 PD patients, 54 (Controls) | ACC: 78% | [47] | |
| ANN | MRI data | 125 (PD patients), 98 (PSP patients), 54 (MSA-P patients), 142 (Controls) | PD ACC: 96.8%, AUC: 0.995 PSP ACC: 93.7%, AUC: 0.982 MSA-P ACC: 95.2%, AUC: 0.990 Controls ACC: 98.4%, AUC: 1.000 |
[48] | |
| Amyotrophic lateral sclerosis | ANN | whole-genome data | 4511 (ALS patients), 7397 (Controls) | ACC: 77% | [49] |
| RF | brain MRI data | 24 (ALS patients), 24 (Controls) | ACC: 80% | [50] | |
| SVM | fMRI data | 32 (ALS patients), 31 (Controls) | ACC: 71% | [51] | |
| Amyloidosis | SVM | proteomics data | 75 (amyloid positive), 78 (Controls) | ACC: 96–99% | [52] |
| Hypophosphatasia | SVM | clinical data | 23 (HPP patients), 22 (Controls) | ACC: 90%, TPR: 87%, TNR: 93%, AUC: 0.936 | [53] |
| Friedreich ataxia | SVM | kinematic data | 30 (Friedreich patients), 14 (Controls) | ACC: 91%, TPR: 90%, TNR: 93%, AUC: 0.91 | [54] |
| Progressive supranuclear palsy | SVM | MRI data | 28 (PSP patients), 28 (PD patients), 28 (Controls) | ACC: 85.8%, TPR: 86%, TNR: 86% PD vs. Controls ACC: 89.1%, TPR: 89.5%, TNR: 89.1% PSP vs. Controls PSP vs. PD 84.7 87.5 83.8 |
[55] |
| Osteogenesis imperfecta | SVM | μCT images | 21 (specimens of 13 OI patients), 19 (specimens from 15 Controls) | AUC: 96% | [56] |
| Mucormycosis disease | ANN | eye photographs | Not clear | ACC: 99.5% | [57] |
| Retinitis pigmentosa | ANN | ultrawide-field images | 150 (RP patients), 223 (Controls) | AUC: 0.998, TPR: 99.3%, TNR: 99.1% of the ultrawide-field pseudocolor group AUC: 1.000, TPR: 100%, TNR: 99.5% of the ultrawide-field autofluorescence |
[58] |
| Cerebral arteriovenous malformation | ANN | Brain CT images | 80 | DSC: 85.2% TPR: 88% TNR: 99% |
[59] |
| Duchenne muscular dystrophy | ANN | cine MRI data | 15 (Duchenne patients), 16 (Controls) | DSC: 97.2 for the sagittal view DSC: 96.1 for the axial view DSC: 96.6 for the coronal view |
[60] |
| Immune thrombocytopenic purpura | ANN | proteomics data | 64 (pITP patients), 70 (sITP patients), 82 (Controls) | TPR: 87.5%, TNR 69.7%, ACC: 75.0% | [61] |
| Multiple osteochondromas | ANN | clinical data | 96 (class I), 137 (class II), 56 (class III) | Class I ACC: 94% Class II ACC: 80% Class III ACC: 79% |
[62] |
| Rett syndrome | ANN | heart rate variation data | 35 (Rett patients), 40 (Controls) | ACC: 88% | [63] |
| Pulmonary arterial hypertension | EL | microRNA expression data | 64 (PAH patients), 43 (Controls) | TPR: 91%, TNR: 64%, ACC: 81%, AUC: 0.85 | [64] |
| Primary sclerosing cholangitis | RF | bacterial 16S rRNA gene sequence data | 24 (PSC patients), 16 (UC patients), 24 (Controls) | PSC AUC: 0.7423 UC AUC: 0.8756 |
[65] |
| Choroideremia | RF | OCT/OCTA data | 16 (eyes with choroideremia), 5 (Controls eyes) | J: 0.876 ± 0.066 | [66] |
| Dermatomyositis | RF | sonographic muscle images | 11 (IBM patients), 19 (DM1 patients), 21 (PM-DM patients) | ACC: 78.4% | [67] |
Performance Metrics Abbreviations: true positive rate (TPR), true negative rate (TNR), positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC), mean reciprocal rank (MRR), Relative Absolute Error (RAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Concordance index (C-index), area under the curve (AUC), Coefficient of determination (R2), Confidence interval (CI), Sørensen–Dice coefficient (DSC), Jaccard similarity index (J).