Table 4 List of the analyzed sample, the CNN algorithm type, tested data (sound or image samples), and their significant findings for the publications that have been surveyed
From: A review on lung disease recognition by acoustic signal analysis with deep learning networks
Study | Method | Splitting strategy | Performance | |||
|---|---|---|---|---|---|---|
Specificity | Sensitivity | Accuracy | Score | |||
VGG16 | Tenfold CV | – | – | 63.09% | – | |
Serbes et al. [144] | SVM | Official 60/40 | – | - | 49.86% | – |
Sen I, et al. [143] | GMM Classifier | – | 90% | 90% | 85.00% | – |
Saraiva et al. [141] | CNN | Random 70/30 | – | – | 74.3% | – |
Yang et al. [172] | ResNet + SE + SA | Official 60/40 | 81.25% | 17.84% | – | 49.55% |
Ma et al. [101] | bi-ResNet | Official 60/40 Random tenfold CV | 69.20% 80.06% | 31.12% 58.54% | 52.79% 67.44% | 50.16% 69.30% |
Pham et al. [130] | CNN-MoE | Official 60/40 Random fivefold CV | 68% 90% | 26% 68% | – | 47% 97% |
Gairola et al. [55] official 60/40 | CNN | Official 60/40 Interpatient 80/20 | 72.3% 83.3% | 40.1% 53.7% | – | 56.2% 68.5% |
CNN | Random 75/25 | – | – | 81.62% | – | |
Acharya and Basu [3] | CNN-RNN | interpatient 80/20 | 84.14% | 48.63% | – | 66.38% |
Allahwardi & Altan et al. [14] | Deep Belief Networks (DBN) | – | 93.65% 73.33% | 93.34% 67.22% | 95.84% 70.28% | |
Kochetov et al. [80] | RNN | Interpatient fivefold CV | 73% | 58.4% | – | 65.7% |
Minami et al. [109] | CNN | Official 60/40 | 81% | 28% | – | 54% |
Georgios Petmezas et al. [129] | CNN-LSTM with FL | Interpatient tenfold CV LOOCV | 84.26% – | 52.78% 60.29% | 76.39% 74.57% | 68.52% – |
Chambres et al. [31] | HMM SVM | Official 60/40 | 56.69% 77.80% | 42.32% 48.90% | 49.50% 49.98% | 39.37% 49.86% |
Oweis et al. [122] | ANN | – | 100% | 97.8% | 98.3% | – |
Jakovljevi´c and Lonˇcar-Turukalo [73] | HMM | Official 60/40 | – | – | - | 39.56% |
Bahoura [22] | GMM | – | 92.8% | 43.7% | 80.00% | – |
Emmanouilidou D et al. [46] | RBF SVM Classifier | – | 86.55 (± 0.36) | 86.82 (± 0.42) | 86.70% | – |
ResNet + NL | Official 60/40 Interpatient fivefold CV | 63.20% 64.73% | 41.32% 63.69% | – | 64.21% 52.26% | |
Nangia et al. [21] | CNN | – | – | – | 94.24% | 93.6% |
Pramono RX et al. [4] | SVM | – | 83.86% | 82.06% | 87.18% | 82.67% |
Nguyen and Pernkopf [118] | ResNet | Official 60/40 Official 60/40 | 79.34% 82.46% | 47.37% 37.24% | – 73.69% | 58.29% 64.92% |
Bardou D et al. [23] | CNN | – | – | – | 95.56% | – |
Aykanat M et al. [18] | ANN | – | 86% | 86% | 76.00% | – |
Chamberlain et al. [30] | – | – | 0.56 | – | 86% Wheeze | – |