Automatic Objective Magnification Detection in Brightfield Microscopy by Use of Neural Network

In brightfield microscopy, commonly used digital cameras are still lacking many features that are required for high-quality digital images. Many of those features such as correction for geometric distortions, chromatic aberration, and vignetting are often done automatically by both professional and smartphone cameras. In scientific imaging, those in-camera corrections are even more essential but, in microscopy, they cannot be done automatically because information about the magnification and microscope objective is often not available to the camera. In this paper, we show that real-time artificial neural networks can be successfully used for the detection of objective magnifications in microscopy. We demonstrate that a real-time artificial neural network trained on only 8000 images can successfully classify four different objective magnifications of 4X, 10X, 20X, and 40X with 92% accuracy on previously unseen images. The approach that we put forward may enable manufacturers to implement these features into their microscopy cameras.