Skip to main content

Table 1 Statistic of our data set

From: Real-time estimation of the remaining surgery duration for cataract surgery using deep convolutional neural networks and long short-term memory

Role of the dataset

Main model Pre-training

Transferability Evaluation

Hospital

A

B

C

 

Surgeon 001

Surgeon 002

Surgeon 003

Video Number

2310

256

54

Mean ± standard deviation (s)

318.7 ± 83.4

547.9 ± 146.3

515.5 ± 140.3

Median (s)

300.0

513.0

480.0

Naïve MAE

Mean (s)

54.9 ± 62.9

106.2 ± 100.6

104.6 ± 93.5

Mean (%)

16.6 ± 14.5

19.1 ± 14.3

20.2 ± 16.1

Median (s)

52.4 ± 67.7

102.3 ± 110.3

97.3 ± 107.2

Median (%)

14.8 ± 13.8

17.1 ± 13.3

17.4 ± 15.5

  1. The videos are decoded as 1FPS and pre-processed by the TCN model. The performance of the naïve method is calculated with mean and median (MAE in seconds and its percentage % of video duration)
  2. MAE mean absolute error