Table 10.
Summary of studies for detection of Covid-19 from cough, breath, and sound signals. UC stands for the University of Cambridge and Sen stands for Sensitivity.
| Study |
Dataset |
Method |
Modality |
Acc (%) |
Pre |
Rec/Sen |
F1 |
Spe (%) |
AUC (%) |
|---|---|---|---|---|---|---|---|---|---|
| Studies using datasets different from ours | |||||||||
| Lella ve Pja, 2022 [27] | UC | DCCN | Breath + Cough + Voice | 95.45 | - | - | 96.96 | - | - |
| Imran et al., 2020 [6] | Privately collected | DTL-BC | Cough | 92.85 | 91.43 | 94.57 | 92.97 | 91.14 | - |
| Brown ve arkadaşları, 2020 [30] | UC: University of Cambridge | SVM | Cough | 0.80 | 0.72 | 82.00 | |||
| Mohammed et al., 2021 [39] | Collected from Github | Scratch CNN | Cough | 0.77 | 0.80 | 0.71 | 0.75 | 0.77 | |
| Chaudhari et al., 2020 [90] | Data is collected from Ruijin Hospital | AI-based method | Cough | - | - | - | - | - | 77.1 |
| Hassan et al., 2020 [91] | Data collected from COVID affected 14 patients | Recurrent Neural Network (RNN) | Cough | 97 | - | - | - | - | - |
| Hassan et al., 2020 [91] | Data collected from COVID affected 14 patients | RNN | Breath | 98.2 | - | - | - | - | - |
| Laguarta et al., 2020 [89] |
MIT open voice data set |
CNN based method |
Cough |
98.5 |
- |
98.5 |
- |
94.2 |
97.0 |
|
Studies using the same datasets as ours | |||||||||
| Harvill et al., 2021 [92] | COUGHVID | LSTM | Cough | - | - | - | - | - | 85.35 |
| Hamdi et al., 2022 [11] | COUGHVID | CNN-LSTM | Cough | 91.13 | 90.47 | 90.93 | 90.71 | 91.31 | 91.13 |
| Orlandic et al., 2021 [34] | COUGHVID | eXtreme Gradient Boosting (XGB) | Cough | 88.1 | 95.5 | 80.8 | - | 95.5 | 96.5 |
| Our Model | COUGHVID | CovidCoughNet (CNN based) | Cough | 99.19 | 0.99 | 0.98 | 0.98 | 97.77 | 98.44 |
| Pahar et al., 2021 [19] | Coswara | ResNet50 | Cough | 95.33 | - | 93 | - | 98.00 | 97.6 |
| Tena et al., 2021 [93] | UC + UL + Coswara + Pertussis + Virufy | Random Forest | Cough | 98.79 | 90.97 | 93.81 | 92.10 | 81.54 | 96.04 |
| Xue ve Salim, 2021 [94] | UC + Coswara | Transformer-CP + Transformer-CP | Cough | 84.43 | 84.57 | 73.24 | 78.50 | 90.03 | |
| Chowdhury et al., 2022 [7], | NoCoCoDa + Cambridge + Virufy+ Coswara |
Extra-Trees Algorithm | Cough | - | 1.00 | 0.97 | - | - | 95 |
| Sharma et al., 2020 [32] | Coswara | Random Forest | Breath + Cough + Voice | 66.74 | - | - | - | - | - |
| Our Model | Coswara | CovidCoughNet (CNN based) | Cough | 99.26 | 1.00 | 0.98 | 0.99 | 98.48 | 98.48 |
| Our Model | Coswara | CovidCoughNet (CNN based) | Breathing | 98.52 | 0.98 | 0.98 | 0.98 | 97.92 | 97.92 |
| Our Model | Coswara | CovidCoughNet (CNN based) | Voice (Vowel-a) | 99.63 | 1.00 | 0.99 | 0.99 | 99.24 | 99.24 |