Current　methods　of　evaluating　the　degree　of　diabetic　retinopathy　are　highly　subjective　and　have　no　quantitative　standard.　To　objectively　evaluate　the　slight　changes　in　tissue　structure　during　the　early　stage　of　retinal　diseases,　a　subjective　interpretation　and　qualitative　analysis　of　the　pathological　sections　of　retinal　HE　in　diabetic　animals　is　required　for　screening　and　evaluating　the　degree　of　diabetic　retinopathy　and　drug　efficacy.　To　develop　an　innovative　method　for　screening　and　evaluating　the　degree　of　diabetic　retinopathy　and　drug　treatment　based　on　artificial　intelligence　algorithms.　Based　on　the　change　law　of　the　early　nerve　fiber　layer　and　the　ganglion　cells,　we　get　disparate　characteristics　of　the　microscopic　image　of　diabetes　animal　retina　HE　slices.　Using　image　recognition　and　deep　learning　methods　on　these　HE　slices,　we　can　identify　the　changes　in　the　ganglion　cells　and　nerve　fiber　layer　for　diagnosing　early　retinopathy　and　evaluated　the　therapeutic　effect　of　the　potential　drugs.　We　conduct　quantitative　calculation　per　unit　length　of　the　nerve　fiber　layer　and　total　area　of　the　nerve　fiber　layer　to　identify　biology　significance　of　edema.　Additionally,　we　also　perform　quantitative　calculation　with　the　number　of　unit　area　ganglion　cells　to　identify　the　section　in　biology　cell　hyperplasia.　Finally,　we　get　the　significance　of　quantitative　calculation　on　the　unit　cell　area　to　identify　ganglion　cell　shriveling　in　biology.　In　addition　to　the　evaluation　of　the　disease　degree　and　changes,　we　also　obtained　retinal　HE　sections　after　different　drug　interventions　and　evaluated　the　therapeutic　effect　of　the　drugs.　This　study　presents　a　novel　quantitative　method　for　screening　and　evaluating　of　diabetic　retinopathy　and　drug　efficacy.