This document reviews various methods for detecting microaneurysms in retinal images to grade diabetic retinopathy. It summarizes three methods: 1) A double-ring filter method that uses a filter to detect candidate lesions and removes false positives near blood vessels. 126 image features are extracted and an neural network classifies lesions. 2) A local rotating cross-section profile analysis method that analyzes directional cross-sections centered on local maxima pixels to calculate attributes of peaks, which are used to classify candidates. 3) An ensemble-based framework method that extracts features using multiple classifiers whose results are combined to detect microaneurysms.