Table 8.

Cross database recalls for best performing models trained with different frameworks and SVD database.

Framework	(Feature) task	Hyperfunction	Laryngitis	Leukoplakia	Normal	Reinke’s Edema	Paralysis	Polyp	Average
AVFAD
CaRLab 2025	(MFCCDD) /a/	–	66	–	2	52	71	49	48
	(MFCCDD) inj. P/N	–	66	–	13	39	44	50	42
USVAC 2025 2	(Wav2Vec) ensemble	–	59	–	2	33	32	58	37
	(MFCCDD) inj. P/N	–	55	–	14	12	48	14	29
da Silva Moura 2024	(UnispeechSAT) phrase	–	100	–	0	3	99	5	41
	(MFCCDD) inj. P/N	–	70	–	21	12	73	11	37
narrow	(UnispeechSAT) ensemble	–	86	–	0	1	51	0	28
	(MFCCDD) inj. N	–	50	–	32	5	48	2	27
MEEI
CaRLab 2025	(MFCCDD) /a/	4	–	25	49	50	62	86	46
	(MFCCDD) inj. P/N	7	–	100	78	0	48	64	50
USVAC 2025 2	(MFCCDD) /a/	9	–	75	44	50	43	64	48
	(Wav2Vec) inj. P/N	41	–	0	91	50	24	36	40
da Silva Moura 2024	(MFCCDD) all	3	–	75	84	50	29	86	54
	(UnispeechSAT) inj. P/N	16	–	100	86	100	0	71	62
narrow	(UnispeechSAT) /a/	11	–	0	25	50	76	0	27
	(UnispeechSAT) inj. P/N	64	–	0	73	0	0	36	29
VOICED
CaRLab 2025	(MFCCDD) phrase	–	61	–	2	67	20	67	43
	(Wav2Vec) inj. P/N	–	47	–	56	0	0	33	27
USVAC 2025 2	(Wav2Vec) /a/phrase	–	21	–	30	67	20	17	31
	(Wav2Vec) inj. P/N	–	0	–	75	0	20	0	19
da Silva Moura 2024	(UnispeechSAT) phrase	–	100	–	0	0	100	0	40
	(Wav2Vec) inj. P/N	–	0	–	79	33	20	33	33
narrow	(MFCCDD) inj. P/N	–	0	–	0	0	0	83	17
	(UnispeechSAT) /a/	–	0	–	54	33	80	0	33

Bold values mark the best-performing configuration.