The development of deep learning approaches to detect, segment or classify structures of interest has transformed the field of quantitative microscopy. High-throughput quantitative image analysis presents a challenge due to the complexity of the image content and the difficulty to retrieve precisely annotated datasets. Methods capable of reducing the annotation burden associated with the training of a deep neural network on microscopy images becomes primordial.

Here we introduce a weakly supervised MICRoscopy Analysis neural network (MICRA-Net) that can be trained on a simple main classification task using image-level annotations to solve multiple more complex tasks such as semantic segmentation. MICRA-Net relies on the latent information embedded within a trained model to achieve performances similar to established architectures when no precisely annotated dataset is available. This learnt information is extracted from the network using gradient class activation maps, which are combined to generate detailed feature maps of the biological structures of interest.

We demonstrate how MICRA-Net substantially alleviates the expert annotation process on various microscopy datasets and can be used for high-throughput quantitative analysis of microscopy images.

Read the article in Nature Machine Intelligence