| [1] |
Using fever as parameter for diagnosis increased the sensitivity from 33% to 94% |
The study was carried on a small sample size |
| [32] |
Simplicity to detect features of the cough and low-cost implementation |
Existence of the crackle is not necessary to detect pneumonia and that led to specificity reduction |
| [6] |
Method uses non-contact measurements; therefore, it does not need extensive sterilization procedures |
Increased computational time due to using long windows in the classification |
| [24] |
Improve the accuracy of WHO case management algorithm for pediatric pneumonia. |
High cost and limited number of staff to test Not widely tested |
| [63] |
The SVM was determined to be more accurate than LR in classifying the cough sound signals, particularly with the bio-mimicking features |
– |
| [56] |
Low-cost solution can be implemented in mobile phones |
Noisy sounds for 4.1% of files |
| [4] |
Algorithm compliance with physician diagnosis |
Small sample size. |
| [52] |
Clear sound took from children |
Interpreted missing values in potential predictor variables as an absence of the respective risk factor, which might also have affected the results. However, the number of missing values did not exceed 5.8% in any of the potential predictor variables. |
| [28] |
“Cold air” and/or “talking” as cough triggers |
The exact role of “cold air” and/or “talking” as cough triggers in the pathogenesis of CVA remains unclear. Functional analysis using single photon emission computed tomography or functional magnetic resonance imaging need to be made for validation |
| [83] |
Good detection of different pulmonary diseases |
Small sample size. |
| [84] |
Low cost to be implemented on smartwatch |
Small sample size, and the low accuracy |
| [7] |
Low-cost mobile app |
Unclear classification due to limited features because it is only diagnosing Asthma, COPD, and Allergic Rhinitis, and other related diseases to pulmonary disease are classified under other diseases. |
| [46] |
Low-cost mobile app |
Need personal cough in the enrollment phase |
| [18] |
Low-cost mobile app |
Noisy sound |
| [59] |
Low-cost mobile app |
Small sample size |
| [85] |
Low-cost mobile app |
low accuracy |
| [33] |
– |
Small sample size |
| [14] |
Convenient, and easy to apply |
Small sample size |
| [37] |
Good detection |
Slow performance due to huge dataset |
| [76] |
Low-cost system |
Slow performance due to huge dataset |
| [62] |
Low-cost mobile app |
Noisy image |
| [25] |
Increased accuracy due to utilizing genetic algorithm (GA) and multilayer neural networks (MLNNs) |
High time consumption |
| [55] |
Low-cost mobile app |
Low performance |
| [78] |
Low-cost solution |
Small sample size |
| [50] |
Low cost and can be applied in mobile |
No update for the dataset online |
| [51] |
Low-cost |
No update for the dataset online |
| [35] |
Low-cost |
– |
| [21] |
– |
Low performance |
| [58] |
High accuracy |
Small dataset size and high complexity system |
| [13] |
Low cost |
Small dataset size |
| [60] |
Low cost |
Number of features affect the performance and battery consumption |
| [48] |
Achieved overall good cough detection capability and noise robustness |
– |
| [69] |
High sensitivity in spirometer |
Number of features affect the performance and battery consumption |
| [2] |
High performance |
Low number of features in classification |
| [39] |
Good performance |
Low quality sound |
| [8] |
Low complexity and good performance |
Small dataset size |
| [75] |
Simultaneous implementation with other potential technologies such as microwave imaging and ultrasound imaging that may be capable of detecting consolidations and mucus in lungs |
70% of the coughs were dry, so it degrades the classification accuracy |
| [29] |
Good accuracy |
Unclear voice samples due to low processing capabilities of the Raspberry Pi |
| [5] |
More specificity due to the hybrid model MFCC + SVM |
Slow performance |
| [42] |
Good detection performance |
Low energy cough signals producing a lower detection rate |
| [88] |
Low-cost mobile app |
Small dataset size, the efficiency is dependent on the dataset size, and no real time update on the database |
| [77] |
Low-cost mobile app |
Highly affected by the background noise |
| High accuracy |
Small dataset size |
| [64] |
Different recording devices system |
Low quality sound |
| [68] |
Good detection |
Unmentioned sample size |
| [71] |
Good detection |
Feature extraction problems |
| [81] |
Low-cost mobile app |
No accuracy results mentioned |
| [89] |
Low-cost mobile app |
– |
