Comparing the Clinical Viability of Automated Fundus Image Segmentation Methods

View full-text article in PMC

. 2022 Nov 23;22(23):9101. doi: 10.3390/s22239101

Algorithm A1 Copeland score calculation.

Require:
Survey data where each entry is a tuple $(c a n d i d a t e_{1}$ , $c a n d i d a t e_{2}$ , $w i n n i n g_c a n d i d a t e)$ . Each candidate is one of the segmentation masks involved in the survey. I $w i n n i n g_c a n d i d a t e$ is one of the $c a n d i d a t e_{1}$ or $c a n d i d a t e_{2}$ images.

% An algorithm to calculate the Copeland score based on candidate pairwise comparisons %
1:
function calculateCopelandScore(survey_data)
2:
$c a n d i d a t e s \leftarrow$ eight segmentation masks for the same fundus image
3:
$c a n d i d a t e_p a i r s \leftarrow g e t C o m b i n a t i o n s (c a n d i d a t e s)$
4:
5:
$r a n k i n g_t a b l e \leftarrow e m p t y_t a b l e$
6:
for $c a n d i d a t e$ in $c a n d i d a t e s$ do
7:
$c o p e l a n d_s c o r e \leftarrow 0$
8:
$r a n k i n g_t a b l e . i n s e r t (c a n d i d a t e, c o p e l a n d_s c o r e)$
9:
end for
10:
11:
for $(c_{1}, c_{2})$ in $c a n d i d a t e_p a i r s$ do
12:
$c_{1}_w i n s \leftarrow c o u n t W i n s (c a n d i d a t e = c_{1})$
13:
$c_{2}_w i n s \leftarrow c o u n t W i n s (c a n d i d a t e = c_{2})$
14:
if $c_{1}_w i n s > c_{2}_w i n s$ then
15:
$r a n k i n g_t a b l e . w h e r e (c a n d i d a t e = c_{1}) . c o p e l a n d_s c o r e + = 1$
16:
else if $c_{1}_w i n s < c_{2}_w i n s$ then
17:
$r a n k i n g_t a b l e . w h e r e (c a n d i d a t e = c_{1}) . c o p e l a n d_s c o r e - = 1$
18:
end if
19:
end for
20:
end function

% Counts how many times has $c a n d i d a t e$ won in pairwise comparisons according to survey data stored in $s u r v e y_e n t r i e s$ . %
21:
functioncountWins(survey_entries, candidate)
22:
$c o u n t e r \leftarrow 0$
23:
for $e n t r y$ in $s u r v e y_e n t r i e s$ do
24:
$c_{1} \leftarrow e n t r y . c a n d i d a t e_{1}$
25:
$c_{2} \leftarrow e n t r y . c a n d i d a t e_{2}$
26:
if $c a n d i d a t e$ in ( $c_{1}, c_{2}$ ) then
27:
if $c a n d i d a t e$ is $e n t r y . w i n n i n g_c a n d i d a t e$ then
28:
$c o u n t e r + = 1$
29:
end if
30:
end if
31:
end for
32:
return $c o u n t e r$
33:
end function

% Get all combinations of unique candidate pairs. That is 24 for 8 segmentation images. %
34:
functiongetCombinations(candidates)
35:
$n \leftarrow 8$
36:
$c o m b i n a t i o n s \leftarrow e m p t y_l i s t$
37:
for i in $(0$ to $n)$ do
38:
$c_{1} \leftarrow c a n d i d a t e s . a t_p o s i t i o n (i)$
39:
for j in $(0$ to $n)$ do
40:
$c_{2} \leftarrow c a n d i d a t e s . a t_p o s i t i o n (j)$
41:
$c o m b i n a t i o n s . i n s e r t ((c_{1}, c_{2}))$
42:
end for
43:
end for
44:
end function