Skip to main content
. 2021 Aug 16;9(9):E1361–E1370. doi: 10.1055/a-1507-4980

Table 1. Number of images used for training, validation, and testing in both considered splits.

Random split Patient split Total
Training (after augmentation) Validation Testing Training (after augmentation.) Validation Testing
Small bowel Normal 712
(5,696)
101 204 714
(5,712)
101 202 1017
Normal with bubbles/debris 1415
(11,320)
202 406 1,429
(11,432)
202 392 2023
Lymphoid hyperplasia 32
(256)
4 10 32
(256)
4 10 46
Non-ulcerated inflammation 21
(168)
2 6 22
(176)
2 5 29
Aphthous ulceration 514
(4,112)
73 148 538
(4,304)
73 124 735
Ulcer 504
(4,032)
72 144 520
(4,160)
72 128 720
Fissure/large ulcer 280
(2,240)
40 82 285
(2,280)
40 77 402
Colon Normal 150
(1,200)
21 44 154
(1,232)
21 40 215
Normal with bubbles/debris 901
(7,208)
128 258 916
(7,328)
128 243 1287
Non-ulcerated inflammation 266
(2,128)
37 76 270
(2,160)
37 72 379
Aphthous ulceration 184
(1,472)
26 54 193
(1,544)
26 45 264
Ulcer 237
(1,896)
33 68 254
(2,032)
33 51 338
Fissure/large ulcer 203
(1,624)
28 58 219
(1,752)
28 42 289
Total 5,419
(43,352)
767 1,558 5,546
(44,368)
767 1,431 7744