Table 1.
Overview of papers about preoperative assessment included in our analysis
| Author, years | Study design | Objective | Final cohort | Outcomes | Type of ML | Prediction performance |
|---|---|---|---|---|---|---|
| Zhou CM 2021 | Retrospective single center | Prediction of difficult tracheal intubation in obese patients using six approaches from various ML fields | 1015 | Prediction of difficult tracheal intubation | LR, TR, RF, Gbdt, Xgbc, Gbm | Training vs testing group: LR AUC 0,68–0,70; TR AUC 0,71–0,60; RF AUC 0,92–0,58; Gbdt AUC 0,78–0,71; Xgbc AUC 0,73–0,71; Gbm AUC 0,81–0,66 |
| Mencar C 2020 | Observational multicentric | Efficacy and clinical applicability of different ML methods based on demographic information and questionnaire data to predict OSA severity | 313 | Prediction of obstructive sleep apnea syndrome severity | MV, NB, k-NN, Ct, RF, SVM AdaBoost-SVM, CN2 rule induction, ML, LR, k-NN, RT, SVR, AdaBoost-SVR | SVM AUC 0,65–0,61 RF AUC 0,63 |
| Pépin JL 2020 | Prospective observational single center | Evaluation of mandibular movement monitoring during sleep coupled with an automated analysis by ML for OSA diagnosis | 376 | OSA diagnosis | Sr RDI | Sr-RDI ≥ 5 events/h AUC 0,95; PSG-RDI ≥ 15 events/h AUC 0,93 |
| Keshavarz Z 2020 | Retrospective, single center | Development of a model for predicting OSA to select the best model to determine and screen high-risk OSA patients | 231 | OSA diagnosis | NN, NB, LR, KNN, SVM, RF | NN AUC 0.75; NB AUC 0,76; LR AUC 0,76; KNN AUC 0.65; SVM AUC 0.72; RF AUC 0.75 |
| Gao WD 2019 | Retrospective | Detection of OSA extracting the features of the heartbeat interval signal and the respiratory signal | N/A | OSA diagnosis | Model fusion (LR-SVM) | Sensitivity 74%, specificity 75%, accuracy 75% |
| Tiron R 2020 | Prospective, single center | Determining of sleep and breathing patterns, and then analyzing results to track sleep-related health risks associated with sleep apnea | 248 | Performance of the Firefly technology as a screener for a clinical threshold of apnea hypopnea index ≥ 15 | Firefly technology | ROC AUC (training 0.95, test 0.92); PR AUC (training 0.87, test 0.89) |
| Cheng Q 2017 | Prospective, single center | Predicting pulmonary function by improved classification models with sole inputs being motion sensors from carried phones | 35 | To categorize patients into the correct GOLD stage | SVM | Accuracy 99% |
| Viswanath V 2018 | Prospective, multicenter | Performing a spirometry test using only the audio data from the microphone of a standard smartphone providing automatic feedback | 20505 | Pulmonary function | NB, k-NN, Log Reg (L1) Log Reg (L2), RF, Gradient Boosting VGG CNN Gated-CRNN | Mel spectogram Naive Bayes precision 0.80; Mel spectogram k-NN precision 0.94; Mel spectogram Log Reg (L1) precision 0.94; Mel spectogram Log Reg (L2) precision 0.93; Mel spectogram RF precision 0.96; Mel spectogram Gradient Boosting precision 0.96; Mel spectogram VGG CNN precision 0.97; Mel spectogram Gated-CRNN precision 0.98 |
| Assaf D 2021 | Retrospective, single center | To improve preoperative diagnosis of hiatal hernia in patients candidates for BS | 2482 | Diagnosis of hiatal hernia | ML decision tree model | Achieving 38.5% sensitivity and 92.9% specificity, ML models increased sensitivity up to 60.2% compared to swallow study prediction |
LR logistic regression, TR decision tree, RF random forest, Gbdt gradient boosting decision tree, Xgbc extreme gradient boosting, Gbm light GBM, MV majority vote, NB Naive Bayes, k-NN k-nearest neighbor, Ct classification tree, SVM support vector machine, AdaBoost-SVM adaptive boosting SVM, ML machine learning, RT regression tree, SVR support vector regression, AdaBoost-SVR adaptive boosting SVR, Sr RDI sunrise system-derived respiratory disturbance index, OSA obstructive sleep apnea, GOLD Global Initiative for Chronic Obstructive Lung Disease, NN neural network, NB Naïve Bayes, ROC receiver operating characteristics, PR precision recall, ANNs artificial neural networks, LDA linear discriminant analysis, QDA quadratic discriminant analysis, MLP multilayer perceptron, AdaBoost LR adaptive boosting LR, CNN convolutional neural network, RNN recurrent neural network, XGBs gradient boosting machines, OSA obstructive sleep apnea, BS bariatric surgery