Table 1:

3DCNN parameters and in vitro findings from contrast-enhanced ultrasound (US) imaging of a tissue-mimicking flow phantoms.

input	kernel	Layer 1	Layer 2	Layer 3	Layer 4	Accuracy
3 × 3 × 5	3×3×3	8	16	128	-	84.6%
3 × 3 × 5	3×3×3	16	32	128	-	87.0%
5 × 5 × 9	3×3×3	8	16	16	128	85.0%
7 × 7 × 9	3×3×3	8	16	16	128	88.5%
9 × 9 × 9	3×3×3	8	16	16	128	98.1%
9 × 9 × 9	3×3×3	8	16	16	64	89.2%
9 × 9 × 9	3×3×3	8	8	16	64	88.8%
9 × 9 × 9	3×3×3	8	8	8	64	90.6%
9 × 9 × 9	3×3×1	8	16	16	128	88.1%
11 × 11 × 7	3×3×1	800	800	-	-	97.6%