Table 2 Comparing our baselines results on the cataract-101 dataset with phase recognition approaches by other authors in the last 10 years. #P = number of phases used for segmentation. NR = Not reported by the author. Best result in bold font, second best is underlined.

From: Phase recognition in manual Small-Incision cataract surgery with MS-TCN + + on the novel SICS-105 dataset

Author

#P

Accuracy

ROC AUC

PR AUC

Sensitivity

Specificity

MS-TCN++ (ours) on Cataract-101

10

0.913

0.994

0.865

0.912

0.978

Touma, 202247

10

0.959

NR

0.855

0.610

0.962

Nespolo, 202248

4

NR

0.961

0.960

NR

NR

Mahmoud, 202349

10

0.990

0.940

NR

0.890

NR

Fang, 202250

10

0.965

NR

NR

0.952

NR