A bird’s-eye view of deep learning in bioimage analysis

Abstract

Deep learning of artificial neural networks has become the de facto standard approach to solving data analysis problems in virtually all fields of science and engineering. Also in biology and medicine, deep learning technologies are fundamentally transforming how we acquire, process, analyze, and interpret data, with potentially far-reaching consequences for healthcare. In this mini-review, we take a bird’s-eye view at the past, present, and future developments of deep learning, starting from science at large, to biomedical imaging, and bioimage analysis in particular.

Next-Generation Digital Histopathology of the Tumor Microenvironment

“…A big fundamental improvement step in recent years elevating the next-generation digital pathology approach is the integration of artificial intelligence (AI) algorithms for pattern recognition into the image analysis/image cytometry process [ 44 ]. Over the past few years, these AI tools have become more robust, and with only minimal user input can be applied to automatically detect objects such as nuclei and specific structures as well as for the classification of various anatomical tissue entities within an entire digitized slide [ 44 , 45 ].…”

“…Understanding molecular and cellular interdependencies quickly leads to complex questions, which require the elaboration of extensive algorithms and enormous amounts of computing power to get to an answer. While machine learning has been known to have great potential in this field for many decades, in the recent past it has advanced greatly in its practical use due to the availability of powerful computer technology, in particular parallel computing on multiple CPUs and/or CPU cores as well as due to new software tools, programming languages, and advanced machine learning techniques, which have made the technologies much easier to use without the requirement of advanced theoretical knowledge [ 44 , 45 ]. By engaging state-of-the-art technologies, computer scientists and engineers try to generate models that can provide answers to the complex problems given by nature.…”

“…Machine learning models are generated by learning on examples consisting in observations. Current techniques of machine learning comprise supervised, unsupervised, transfer, federated, and reinforcement learning [ 45 ].…”

Highlights from the 2016-2020 NEUBIAS training schools for Bioimage Analysts: a success story and key asset for analysts and life scientists

“…During the unrolling of the programme there were shifts in the relative perceived importance of topics (seen also in the applications). For example, MATLAB became less requested and BIA with Python libraries and tools became a common request, which coincided with a significant increase in publication of machine/deep learning applications to BIA ( Meijering 2020 ). Handling of “big-data” appeared also on increasing demand, and both topics became prominent in TS after 2017-2018.…”

SplineDist: Automated Cell Segmentation With Spline Curves

“…Due to the sheer amount of imaging data acquired in modern biology experiments and because segmentation is so essential, the development of automated computational segmentation methods has been at the heart of bioimage analysis since its early days [4]. In the past decade, deep learning approaches have revolutionized this field of research [5]. Convolutional neural networks (CNN) in general, and the U-net model in particular [6], have demonstrated an unprecedented capability to consistently produce excellent segmentation results in a wide range of data, massively reducing the need for manual intervention.…”