(EN) A method 100 of modifying a retina fundus image for a deep learning model is disclosed. In a described embodiment, the method 100 includes converting a retina fundus image 111 to a binary image 131 by converting pixels 172 of the retina fundus image 111 to low intensity modified pixels and high intensity modified pixels of the 0 binary image 131, and determining a first boundary 151 between the low intensity modified pixels and high intensity modified pixels of the binary image 131. The method further includes removing outlier boundary values 162 from the first boundary 151, constructing a second boundary 161 from remaining boundary values 163, identifying the pixels 172 of the retina fundus image that are within the second boundary 161, and 5 constructing a modified retina fundus image 171 containing the identified pixels 172 for a deep learning model.