Vision impairment and vision deteriorating diseases caused by diabetic condition is a growing problem around the world. Early detection of Diabetic Retinopathy(DR) can help in taking pro-active measures in mitigate this problem. If medical interventions are made within a reasonable time-frame, early stages of DR is curable. Image segmentation in medical image processing opened a new dimension for early detection of several chronic diseases and better localization of abnormal growth of tissue or blood vessels. This study discusses classification of DR stages based on four types of DR lesion segmentation, this approach addresses the issue of accurate localization of DR lesions and determining the extent of lesion growth. Lesions patches like Microaneurysms(MI) and Soft Exudates(SE) are difficult to segment because of less prominent spatial information. This study proposes an lesion segmentation architecture for segmentation of low pixel intensity features like MI and SE along with other two categories, Hemorrhages(HA) and Hard Exudates(HE). Advancements in deep learning techniques, which enabled computer aided systems to surpass human capabilities. Mass screening of populations with risk of DR can be done along with lesion growth assessment. Proposed lesion quadrant classification algorithm(LQCA) incorporates four binary segmentation models, segmentation of DR lesions and information about lesion extent helps in staging DR condition more accurately.