The early detection of retinal microaneurysms significantly contributes to the prevention of visual impairment in individuals with diabetic retinopathy. During a fundus examination, this research describes a unique method for locating retinal microaneurysms. The strategy being discussed uses a circular frame of reference and radial gradient-based features. Additionally, the Boolean operators ‘AND’ and ‘OR’ may be used to interchange the set operations of intersection and union. Dilation is a kind of geometric transformation in which a figure’s size is changed, but the shape and proportions are kept the same. The first step in this approach involves preprocessing the fundus images and detecting potential microaneurysm candidates using morphological processing and adaptive thresholding algorithms. This work introduces two methods for extracting features, namely, circular reference-based shape highlights (CR-SF) and spiral gradient-based surface highlights (RG-TF), which aim to differentiate between microaneurysms and non-micro aneurysms. The robust backpropagation with PCA balanced data analysis machine learning technique is used to analyse the surface highlights that were retrieved from the candidates in terms of colour, shape, and appearance. During the testing phase, the extracted features from the test image are used to categorise the presence of microaneurysms and non-microaneurysms. The evaluation of the test was conducted using four unique datasets: MESSIDOR, DiaretDB1, e-ophtha-MA, and ROC datasets. The assessment included measuring many statistical parameters, including accuracy, sensitivity, specificity, AUC, and time complexity. The proposed technique demonstrates high levels of precision, sensitivity, specificity, and AUC, with values of 98.01%, 98.74%, 97.12%, and 0.9172, respectively. The evaluation result suggests that the proposed method for calculating the location of retinal microaneurysms surpasses traditional computation approaches.