. 2023 Jun 27;11:1205009. doi: 10.3389/fbioe.2023.1205009

TABLE 6.

Outcome metrics and test performance.

Article	Classifier	Test performance outcome metrics
Article	Classifier	Acc	Sn/Rc	Sp	PPV/Pc	NPV	AUC	Others
Frakking et al. (2022)	SVM	98	89	100	100	100	-	F1: 0.94
Lee et al. (2006) ⁽ ^a ⁾	RBF	82.1	74.7	87.8	-	-	-	Adj. accuracy: 81.3
Lee et al. (2011) ⁽ ^a ⁾	Airway invasion: LDA Euclidean	-	100	49.4	-	-	-	Adj. accuracy: 74.7
	Valleculae BC: LDA Mahalanobis	-	75.5	91.9	-	-	-	Adj. accuracy: 83.7
	Pyriform sinuses BC: LDA w/Mahalanobis	-	81.7	86.8	-	-	-	Adj. accuracy: 84.2
Merey et al. (2012)	LDA w/Euclidean	62.8	50.7	74.9	-	-	-	-
	LDA w/Mahalanobis	60.6	69.8	51.4	-	-	-	-
	SVM linear	62.0	51.5	72.4	-	-	-	-
	SVM RBF	80.6	80.0	81.2	-	-	-	-
	SVM RBF + B2 optimizer	86.9	89.6	92.2	-	-	-	-
Park et al. (2022) ⁽ ^b ⁾	Logistic Regression	68.2	65.7	70.7	69.3	67.8	0.69	F1: 0.67
	Decision Tree	69.0	62.0	76.0	73.3	66.6	0.70	F1: 0.67
	Random Forest	73.7	70.7	76.7	75.7	72.5	0.78	F1: 0.73
	SVM	69.7	71.0	68.3	69.4	70.2	0.68	F1: 0.70
	GMM	66.2	64.7	67.7	66.3	67.5	0.64	F1: 0.64
	XGBoost	74.8	72.7	77.0	76.8	74.8	0.78	F1: 0.74
Sarraf Shirazi et al. (2014)	SVM	86.0	91.0	84.0	-	-	-	-
Sarraf Shirazi et al. (2012)	Classify population: min distance classifier	90.0	-	-	-	-	-	-
Sarraf Shirazi et al. (2012)	Classify swallow: fuzzy k-means	86.4	86.4	86.4	61.5	96.2	-	-
Sejdic et al. (2013) ⁽ ^a ⁾	Bayes	94.6	92.5	95.6	-	-	-	-
Shu et al. (2022)	Naïve Bayes w/AC-GAN	66.38	39.03	74.6	-	-	-	F1: 22.02
	Naïve Bayes w/AC-GAN	66.38	39.03	74.6	-	-	-	MCC: 0.0324
	K-means w/AC-GAN	72.94	12.40	86.41	-	-	-	F1: 13.24
	K-means w/AC-GAN	72.94	12.40	86.41	-	-	-	MCC: −0.0009
	SVM w/AC-GAN	75.02	21.71	86.84	-	-	-	F1: 22.83
	SVM w/AC-GAN	75.02	21.71	86.84	-	-	-	MCC: 0.0938
	ANN w/AC-GAN	71.39	32.84	79.78	-	-	-	F1: 28.75
	ANN w/AC-GAN	71.39	32.84	79.78	-	-	-	MCC: 0.1171

Classifier column: AC-GAN, auxiliary classifier Wasserstein generative adversarial network; BC, bolus clearance; GMM, gaussian mixture model; LDA, linear discriminant analysis; RBF, radial basis function; SVM, support vector machine; XGBoost, Extreme gradient boosting; w/, with. Outcome metrics column: Acc, accuracy; AUC, area under receiver-operating curve; NPV, negative predictive value; Pc, precision; PPV, positive predictive value; Rc, recall; Sn: sensitivity; Sp, spec-ificity.

^{^a}

Classifiers with feature combination of the best accuracy/adjusted accuracy are shown in this table.

^{^b}

Performance for classifying mild/severe dysphagia or aspirated using model trained by acoustics only (without clinical data) is shown in this table.