Table 11.
Preliminary studies applying XAI to time series data
References | # Cit. | Application | Input Data | AI model(s) | XAI method(s) | Dataset(s) |
---|---|---|---|---|---|---|
Sun et al. (2021) | 0 | Mortality prediction | Longitudinal EHR | TCoN | Attention | MIMIC-III Johnson et al. (2016) |
Thorsen-Meyer et al. (2020) | 11 | Mortality prediction | Longitudinal EHR | LSTM | SHAP | Retrospective study |
Shamout et al. (2019) | 39 | Early detection of in-hospital deterioration | HR, HRV, BT, BP, SpO | Bi-LSTM ensemble | Attention | Retrospective study |
Lauritsen et al. (2020) | 124 | Acute critical illness prediction | Longitudinal EHR | TCN | LRP | Retrospective study |
Jiang et al. (2021) | 0 | Hospital readmission prediction | Longitudinal EHR | T-LSTM | Attention | MIMIC-III, New York State MDW |
Luo et al. (2020) | 45 | Disease diagnosis | Longitudinal EHR | HiTANet | Attention | UCI COPD, HF, CKD datasets |
Maweu et al. (2021) | 16 | Arrhythmia detection | ECG | CNN | Grad-CAM | MIT-BIH Arrhythmia database subsetPlawiak (2017) |
Ma et al. (2020) | 0 | AMI detection | MTS | CNN | PL | MIMIC-III |
Bois et al. (2020) | 2 | BGC forecasting | longitudinal BGC, CHO intake, and insulin data | RETAIN Choi et al. (2016) | Attention | IDIAB dataset |
Zhang et al. (2020) | 36 | Epileptic seizure detection | EEG | CNN | Attention | TUH-EEG database Obeid and Picone (2016) |
Li and Sano (2020) | 24 | Stress detection | EDA, ST, 3-D Acc | LC-LSTM-DAE | Attention | Pilot study |
He et al. (2021) | 1 | BP forecasting | Longitudinal EHR | MAC-LSTM | Attention | Retrospective study |