Table 3.

Comparative performance of the graph CNN model in identifying normal versus abnormal frames in the four VCE videos in the testing set.

Metrics	Video 6	Video 7	Video 8	Video 9
Frames	14,173	16,909	10,037	19,104
Segments	770	1124	248	1071
Disease Categories	5	2	3	3
Precision	82.0%	69.8%	33.6%	76.7%
Sensitivity	91.1%	80.4%	60.1%	88.9%
F-score	89.9%	82.2%	57.8%	85.9%
Accuracy	89.9%	84.8%	56.0%	85.9%