Table 1.

The existing DL models on upper airway segmentation and their functions and performance.

Authors	DL Models	Year	Training Dataset	Validation/Test Dataset	Functions	Best Performance of DL	Time- Consuming
Jacobs et al. [28]	3D U-Net	2021	48	25	Segmentation of pharyngeal airway space	Precision: 0.97 ± 0.02 Recall: 0.98 ± 0.01 Accuracy: 1.00 ± 0.00 DSC: 0.98 ± 0.01 IoU: 0.96 ± 0.02 95HD: 0.82 ± 0.41 mm	No
Choi et al. [29]	CNN	2021	73 for segmentation 121 for OSAHS diagnose	15 for segmentation 52 for OSAHS diagnose	Segmentation of upper airway, computational fluid dynamics and OSAHS assessment	Upper airway flow characteristics Accuracy: 0.702 ± 0.048 Sensitivity: 0.893 ± 0.048 Specificity: 0.593 ± 0.053 F1 score: 0.74 ± 0.033 DSC: 0.76 ± 0.041 OSAHS diagnosis Accuracy: 0.815 ± 0.045 Sensitivity: 0.893 ± 0.048 Specificity: 0.862 ± 0.047 F1 score: 0.0876 ± 0.033	6 min
Yuan et al. [30]	CNN	2021	102	21 for validation 31 for test	Segmentation of upper airway	Precision: 0.914 Recall: 0.864 DSC: 0.927 95HD: 8.3	No
Spampinato et al. [31]	CNN	2021	20	20	Segmentation of sinonasal cavity and pharyngeal airway	DSC: 0.8387 Matching percentage: 0.8535 for tolerance 0.5 mm 0.9344 for tolerance 1.0 mm	No
Oz et al. [32]	CNN	2021	214	46 for validation 46 for test	Segmentation of upper airway	DSC: 0.919 IoU: 0.993	No
Lee et al. [33]	Regres-sion Neural Network	2021	243	72	Segmentation of upper airway	r2 = 0.975, p < 0.001	No