Table 2.
Summary of final articles.
| Authors, Year | Research aims | Data resources | Data types | Input | Output | AI methods | AI Performance metrics |
Explainable techniques |
|---|---|---|---|---|---|---|---|---|
| Alahmadi, A. et al., 2021 [29] | To develop an explainable rule-based decision tree classification model to automate the detection of QT-prolongation at risk of Torsades de Pointes (TdP) | Public dataset, clinical trial approved by Food and Drug Administration (FDA) in 2014 | ECG image data | ECG | Classification of Torsade de Pointes (TdP) | Rule-based algorithm | Accuracy Balance Sensitivity Specificity PPV F1-score ROC (AUC) Precision-Recall (AUC) MCC Error rate |
Pseudo-coloring methodology |
| Born, J. et al., 2021 [30] | To develop an explainable classification model for differential COVID-19 diagnosis | Public dataset, Lung Point-Of-Care Ultrasound (POCUS) |
Ultrasound video data | Ultrasound | Classification of COVID-19 | CNN | Precision Recall F1-score Specificity MCC |
CAM |
| Neves, I. et al., 2021 [31] | To develop an explainable ECG classification model on time series | Public dataset, Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) Arrhythmia | ECG image data | ECG | Classification of arrhythmia | KNN CNN |
F1-score Precision Recall AUC |
PFI LIME SHAP |
| Sabol, P. et al., 2020 [32] | To develop an explainable classification model for colorectal cancer diagnosis | Public dataset, Colorectal cancer pathology image |
Histopatho-logical image data | Colorectal cancer pathology image data | Classification of colorectal cancer | CNN | Accuracy Precision Recall F1-score |
CFCMC |
| Tan, W. et al., 2021 [33] | To develop an explainable deep learning model for the automatic diagnosis of fenestral OS | EHR data, the Fudan University | CT scan image data | Temporal bone high-resolution computed tomography (HRCT) | Classification of fenestral otosclerosis | Conventional image processing algorithm | Accuracy Sensitivity Specificity PPV NPV |
Faster-RCNN |
| Derathé, A. et al., 2021 [34] | To explain the previously developed prediction model for surgical practice quality | EHR data, the CHU Grenoble Alpes Hospital | Laparoscopic sleeve gastrectomy (LSG) operation video data | Laparoscopic operation videos | Extraction of the most important variables to predict the quality of surgical practice | SVM | Accuracy Sensitivity Specificity | Value- permutation and Feature-object semantics |