Table 1. Distribution of training data with 90 percent hybrid images.

View	1. Number of original training images	2. Number (percent) of donors	3. Number of times each donor is sampled	4. Number of hybrid images generated	5. Number (percent) of original training images sampled	6. Total number of training images	7. Percent of hybrid images in total training data
3VT	3193	297 (9.3)	22	6534	700 (22)	7234	90.3
3VV	6178	439 (7.1)	15	6585	700 (11)	7285	90.4
LVOT	6735	469 (7.0)	14	6566	700 (10)	7266	90.4
A4C	6029	570 (9.5)	12	6840	700 (12)	7540	90.7
ABDO	5206	475 (9.1)	14	6650	700 (13)	7350	90.5
NT	25082	633 (2.5)	50	31650	3500 (14)	35150	90.0
TOTAL	52423	2883 (5.5)	N/A	64825	7000 (13)	71825	90.2

Distribution of training data where 90 percent of training data consists of hybrid images generated in an offline manner (columns 6, 7). Only about five percent (column 2) of the original training images (column 1) are used to generate all the new hybrid training images (column 4). Overall, the entire hybrid training dataset (column 6) was derived from only 13 percent of the original images. Column 3 denotes the number of times each donor is samples. A total number is not applicable here.