Table 6 Comparative analysis of experimental results obtained on various models/approaches.
From: Integrating physiological signals for enhanced sleep apnea diagnosis with SleepNet
Authors | Proposed method | Accuracy (%) | Specificity (%) | Sensitivity (%) |
|---|---|---|---|---|
Vector valued Gaussian processes | 82.33 | 84.12 | 76.28 | |
Binary classification using two first FBanks, first CC, and first DFA coeff as variables, OSA detection | 84.76 | 81.45 | 86.82 | |
Deep neural network and hidden Markov method | 84.7 | 82.10 | 88.90 | |
Support vector machine (SVM) | 88.20 | 93.9 | 80.00 | |
10-fold cross validation | 88.30 | 88.30 | 88.30 | |
Deep residual network | 83.03 | 85.60 | 78.73 | |
Pre-trained AlexNet | 86.22 | 83.82 | 90.00 | |
The exploration of frequential stacked sparse auto-encoder (FSSAE) and time-dependent cost-sensitive (TDCS) classification approaches was conducted | 85.10 | 84.40 | 86.20 | |
MSDA-1DCNN and weighted-loss time-dependent classification (WLTD) | 89.40 | 89.10 | 89.80 | |
MobileNet V1 with GRU | 90.29 | 90.72 | – | |
CNN-BiGRU with attention | 91.20 | 94.2 | 95.7 | |
CNN + Modified honey badger | 91.30 | 93.60 | 90.10 | |
Multi-layer neural network with discrete wavelet transform | 92.30 | – | – | |
EfficientNet-B2 | 93.33 | – | – | |
Proposed model | Unimodal 1D CNN + BiGRU | 95.08 | 92.12 | 95.67 |
Proposed model | SleepNet (multimodal 1D CNN + BiGRU) | 95.19 | 93.45 | 96.12 |
