TABLE VI.

COMPARISON WITH STATE-OF-ART METHODS ON LOOCV VALIDATION

Dataset	Method	DSC					IOU
		Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5
	UNet (fully)	0.745	0.557	0.791	0.743	0.808	0.674	0.464	0.673	0.617	0.697
	UNet++ (fully)	0.747	0.520	0.771	0.746	0.783	0.679	0.412	0.663	0.632	0.675
	TransUNet (fully)	0.737	0.512	0.757	0.723	0.749	0.663	0.405	0.645	0.601	0.616
	Mean Teacher	0.794	0.600	0.709	0.745	0.752	0.688	0.457	0.576	0.642	0.629
	Deep Co-training	0.742	0.554	0.739	0.745	0.741	0.669	0.459	0.642	0.615	0.622
	Cross Pseudo	0.819	0.659	0.768	0.734	0.770	0.744	0.558	0.668	0.611	0.646
	Duo SegNet	0.748	0.597	0.690	0.729	0.800	0.680	0.494	0.611	0.607	0.688
	MMS (ours)	0.874	0.728	0.843	0.807	0.875	0.784	0.579	0.743	0.675	0.783
Cochlear	Method	MAE					F-measure
		Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5
	UNet (fully)	0.064	0.093	0.052	0.061	0.060	0.795	0.628	0.805	0.775	0.822
	UNet++ (fully)	0.058	0.104	0.061	0.062	0.058	0.792	0.583	0.793	0.768	0.797
	TransUNet (fully)	0.068	0.112	0.064	0.072	0.063	0.753	0.562	0.781	0.734	0.770
	Mean Teacher	0.053	0.088	0.075	0.064	0.062	0.815	0.628	0.731	0.782	0.772
	Deep Co-training	0.064	0.091	0.063	0.067	0.064	0.802	0.630	0.797	0.762	0.767
	Cross Pseudo	0.044	0.079	0.058	0.065	0.059	0.853	0.716	0.801	0.727	0.785
	Duo SegNet	0.062	0.084	0.075	0.068	0.056	0.809	0.662	0.759	0.756	0.815
	MMS (ours)	0.037	0.072	0.046	0.053	0.039	0.879	0.733	0.853	0.806	0.878