From: Use of artificial neural networks in the prognosis of musculoskeletal diseases—a scoping review
Author | Variables used for ANN model | Method to evaluate accuracy | Comparing with other method | Results of ANN accuracy | Platform for ANN modeling |
---|---|---|---|---|---|
Alfieri et al | The gene transcript expression of 190 wound healing, inflammatory, osteogenic, and vascular genes | ROC curve | Better than LASSO LR | AUC = 0.780 | Oncogenomics Online ANN Analysis system |
Almhdie et al | KL grades, OARSI grades, demographic, WOMAC pain, race, and history of knee injury, and TBT | ROC curve | Better than LR | AUCOAI = 0.75; AUCMOST = 0.81 | ImageNet |
Bevevino et al | Demographics, mechanism of injury, wound size and location, fracture types, interval and definitive treatment procedures, rotational or free tissue transfer, skin graft, and neurovascular procedures, and ipsilateral and contralateral orthopaedic injuries | ROC curve | Better than LR and random forest model | AUC = 0.800 (95% CI: 0.770, 0.820) | SAS |
Bowman et al | Demographic, clinical and neurophysiological variablesa | ROC curve | Better than LR | AUC = 0.767 | MATLAB |
Chen et al | Demographic, referral to lower-level medical institutions, urbanization, socioeconomic status, number of comorbidities, intracapsular fracture, hospital level, hospital volume, and surgeon volume | ROC curve | Better than Cox regression model | AUC = 0.930 (95% CI: 0.900, 0.960) | STATISTICA |
Eller-Vainicher et al | Menopausal age, number of pregnancies, breast feeding, smoking habits, alcohol consumption, previous clinical fragility fractures at spine, ribs, wrist and hip, BMI, calcium intake, co-morbiditiesb | ROC curve | Better than LR | AUCSDI≥1 = 0.714; AUCSDI≥5 = 0.823 | TWIST system |
Jalali et al | Demographics, intra/postoperative transfusion, intraoperative surgical and anesthetic management, postoperative management and laboratory results, the occurrence of prespecified intraoperative and postoperative complications, and the length of intensive care unit and hospital stay in calendar days | ROC curve | Inferior than GBM | AUC = 0.790; | Python |
Kim et al. (Korea) | T2-sagittal consecutive lumbar spine MR images, T2-weighted sagittal lumbar spine MR slices, MR images obtained prior to the TFESI | ROC curve | - | AUC = 0.827(95% CI, 0.774,0.909) | Python |
Kim et al. (USA) | Demographic, diabetes, smoking, steroid use, coagulopathy, functional status, ASA class O3, BMI, pulmonary comorbidities, and cardiac comorbidities | ROC curve | Better than LR | AUCcardiac = 0.768; AUCVTE = 0.542; AUCwound = 0.606; AUCmortality = 0.844 | MATLAB |
Lu et al | Age, sex, body mass index, type of sports participation, clinically documented ligamentous laxity, clinical history of instability, radiographic findings, management, recurrent instability, and development of clinically symptomatic osteoarthritis | ROC curve | Inferior than XGBoost | AUCrecurrence = 0.823 (95% CI, 0.821,0.824); AUCsurgery = 0.689 (95% CI, 0.687,0.692); AUCosteoarthritis = 0.692 (95% CI, 0.687,0.697) | R |
Miyoshi et al | Demographic, ESR, TEN, ALB, MONO, RBC, PSL, MTX, HbA1c and Pre bio | ROC curve | - | AUC = 0.750 | WEKA software package |
Norgeot et al | Prior CDAI score, ESR and CRP level, DMARDs, oral and injected glucocorticoids, autoantibodies, age, sex, and race/ethnicity | ROC curve | - | AUCUH = 0.910 (95% CI, 0.860,0.960); AUCSNH = 0.740 (95% CI, 0.650,0.830) | Github |
Salgueiro et al | MPQ, the HAQ-DI, and the anxiety subscale of HADS | ROC curve | Better than LR | AUCANN1 = 0.917; AUCANN2 = 0.947 | - |
Scheer et al | Demographic, radiographic, and surgical factors | ROC curve | - | AUC = 0.890 | SPSS Modeler |
Shin et al | Changes in clinical symptoms and blood inflammatory markers | ROC curve | - | AUC = 0.902 (95%CI, 0.804, 0.999) | Keras and TensorFlow |
Su et al | Sex, age, BMI, region, morning stiffness time, step count, and osteophyte area | ROC curve | - | AUCworse = 0.814; AUCunchanged = 0.700; AUCimproved = 0.761 | R |
Wang et al | Age, EO, PLT, WBC, NEUT, U-SG, U-WBC, U-WBC | ROC curve | - | AUCILD = 0.792; AUCPF = 0.751 | STATISTICA |
Yahara et al | Frontal view of the total spine radiographs: the C7 vertebra and diaphragm; diaphragm and ilium; and C7 vertebra and ilium | ROC curve | - | AUC = 0.700 | MATLAB |