For many applications, such as liquid-liquid or gas-liquid reactions, the generation of monodisperse droplets is of major interest. Therefore, knowledge about the physics of droplet formation is essential and the subject of numerous studies. Droplet formation is usually investigated using optical cameras, which makes optical accessibility necessary. Furthermore, properties defining droplet evolution is obtained from 2D images. In this work, we present a methodology for the 3D investigation of droplet formation in the laminar regime using micro-computed tomography. A special imaging concept and image processing, incorporating the use of a convolutional neural network, is presented. Water droplets are injected into a continuous polydimethylsiloxane stream in a coflowing configuration using a cannula with an inner diameter di = 800 μm and an outer diameter do = 1050 μm that is centered in a thin polymer tube with an inner diameter di = 1600 μm. Volume flow rates of polydimethylsiloxane and water are varied between 0.2 and 0.3 mL min−1. Furthermore, the influence of cannula positioning on droplet formation is investigated. Different quantitative 3D properties are extracted from the CT scans, such as droplet volume and surface of the interface. Thereby, different stages of droplet formation can be identified and the physical understanding of droplet formation is improved.