Algorithm 1 Heterogeneous region extraction

Input: Ml (the source lesion matrix with size m × n.)

Output: SR (the subregion mask matrix list.)

0: function HeterRegionEx(Ml)

0: SR ← null, L ← null, T SR ← null

for p in Ml do

0: L.append((x, y, p))/*x and y are the location*/

0: end for

1: (C1, C2, ···, Ck) ← clustering with elbow rule on data L

2: T SR.append(C1, C2, ···, Ck)

for Ci in T SR do

if Ci is not a connectivity area then

3: (Ci1, Ci2, ···, Cir) ← splitting r connective clusters on Ci

4: T SR.append(Ci1, Ci2, ···, Cir)

T SR.remove(Ci)

end if

0: if size of Ci > 10 then

0: SR.append(Ci)

0: end if

0: end for

end function=0