Table 1 Comparison of performances of automated lesion segmentation algorithm by Mulder et al.27 vs. algorithm of the present study on open data repository of Mulder et al.27 Observer 1 vs. observer 2 refer to lesion masks produced by each of the two manual tracers from their study.

From: Deep learning-based automated lesion segmentation on mouse stroke magnetic resonance images

Age (months)

Time between infarct induction and MRI

Mulder et al. method

Present study

Observer 1

Observer 2

Observer 1

Observer 2

3

18 h (n=6)

0.88 ± 0.03

0.89 ± 0.03

0.789 ± 0.05

(0.74 - 0.87)

0.81 ± 0.05

(0.77–0.90)

4 d (n=3)

0.85 ± 0.01

0.87 ± 0.01

0.68 ± 0.034

(0.63 - 0.72)

0.712 ± 0.03

(0.69–0.75)

1, 12

48 h (n=10)

0.86 ± 0.07

0.85 ± 0.07

0.80 ± 0.06

(0.66 - 0.87)

0.76 ± 0.08

(0.64–0.87)

Average

 

0.86

0.87

0.76

0.76

  1. The mean dice coefficient across animals within each group and the average of the mean dice coefficient across three groups are shown. For the present study, errors refer to standard deviation. For Mulder et al.,2 the type of error was not reported. We additionally report the range of dice scores with parenthesis.
Back to article page