Table 5.
List of XAI studies with tabular data performing clinical validation of explanations
References | # Cit. | Application | Input Data | AI model(s) | XAI method(s) | Dataset(s) |
---|---|---|---|---|---|---|
Beebe-Wang et al. (2021) | 7 | Imminent dementia diagnosis | EHR | XGBoost | SHAP | ROSMAP dataset Bennett et al. (2012) |
Sha et al. (2021) | 4 | AD diagnosis | Plasma metabolite concentrations | Evolutionary algorithms | LGP | AD metabolomic dataset Wang et al. (2014) |
Kim et al. (2021) | 3 | END detection in stroke patients | EHR | LightGBM | SHAP | K-Attention dataset Jung et al. (2019) |
Rashed-Al-Mahfuz et al. (2021) | 6 | CKD diagnosis | EHR | RF, GBDT, XGBoost | SHAP | UCI CKD dataset Rubini and Eswaran (2015) |
Pang et al. (2019) | 10 | Early childhood obesity prediction | EHR | XGBoost | SHAP | Retrospective study |
Zeng et al. (2021) | 5 | Post-operative complication risk prediction | EHR | XGBoost | SHAP | Retrospective study |
Zhang et al. (2021) | 10 | post-LT AKI prediction | EHR | RF | SHAP | Retrospective study |
Lu et al. (2020) | 5 | COVID-19 diagnosis and prognosis | EHR | GBDT, RF | SHAP | Retrospective study |
Gupta et al. (2021) | 1 | COVID-19 recovered subject detection | HR/HRV features | XGBoost | SHAP | Retrospective study |
Pal and Sankarasubbu (2021) | 50 | COVID-19 diagnosis | Cough sounds, symptoms metadata | TabNet, DNN | Attention | Pilot study |
Dissanayake et al. (2020) | 18 | Heart anomaly detection | PCG features | stacked LSTM-CNN-MLP network | SHAP, occlusion maps | PhysioNet DB Goldberger et al. (2000) |