Table 2.

Synthetic breathing artefact data classification results for mean accuracy (A), precision (P), recall (R) and balanced accuracy (BA) results. A 10-fold cross validation was used and each image was labelled once over all folds and standard deviation over folds is reported (mean ± std). All results are multiplied by 1000 and the bold font highlights the best results.

Methods	A	P	R	BA
K-Nearest Neighbours	718 ± 33	724 ± 37	721 ± 30	723 ± 36
Linear SVM	740 ± 41	737 ± 80	744 ± 48	741 ± 76
Decision Tree	707 ± 55	708 ± 42	713 ± 37	711 ± 48
Random Forests	764 ± 56	776 ± 64	781 ± 68	778 ± 61
Adaboost	768 ± 39	768 ± 54	772 ± 50	770 ± 57
Naive Bayesian	788 ± 70	790 ± 43	797 ± 68	793 ± 42
Variance of Laplacian	809 ± 43	820 ± 69	824 ± 55	822 ± 59
NIQE	897 ± 59	899 ± 71	904 ± 61	902 ± 50
Lorch et al. (2017)	896 ± 62	895 ± 47	897 ± 38	896 ± 77
3D CNN	953 ± 89	951 ± 91	961 ± 82	955 ± 70
LRCN	961 ± 41	962 ± 29	964 ± 51	963 ± 30