Deep Learning - Pushing the Frontiers in Medicine and Life Sciences

MetaSystems combines more than 30 years of experience with a constant drive for innovation to provide the best possible solutions for automated imaging.

Deep Neural Networks (DNNs) open new horizons in medicine and life sciences.
DNNs solve sophisticated computer vision tasks, such as object detection and image classification.
DNNs are self-learning statistical models.
MetaSystems applies a supervised learning strategy.
MetaSystems implemented DNNs in different imaging applications.

Deep Neural Networks (DNNs) are a class of artificial intelligence algorithms that solve challenging image processing tasks such as object recognition and image classification by learning from large amounts of sample data. The process of training a DNN is referred to as Deep Learning. The principles of Deep Learning are decades old, but the relatively recent availability of big data and computing power has helped DNNs achieve a breakthrough. MetaSystems has implemented algorithms based on Deep Neural Networks (DNNs) in various imaging applications in medicine and life sciences.

Deep Learning Uses Deep Neural Networks (DNN)

Deep learning is an advance in the area of artificial intelligence (AI). Prominent examples of deep learning are facial recognition in smartphone cameras and the understanding of speech. Deep learning has great potential for applications in medicine and life sciences as well.

MetaSystems has implemented software algorithms based on Deep Neural Networks (DNNs), whose architecture is modeled on the network of neurons in the human brain. More precisely, DNNs are large statistical models with millions of adaptable parameters. The huge number of parameters enables a DNN to learn abstract features for image differentiation by itself.

Supervised Learning

During the learning phase of the DNN, the prediction made by the DNN for each training image is compared to the actual, correct output (ground-truth). The DNN can measure its own accuracy and adjusts the parameters for classification accordingly. The source for this supervised learning are images labeled by an expert. However, if acquired microscopy images drastically differ from those used during the self-learning phase of the DNN, it may be necessary to train a different DNN.

MetaSystems offers innovative solutions for automated microscopy imaging for numerous applications with brightfield and fluorescence illumination. In cooperation with global partners, MetaSystems has developed DNNs for different imaging applications.

Learn more about our DNN based applications.

Contact Us