Imaging in focus: An introduction to denoising bioimages in the era of deep learning.

Fluorescence microscopy enables the direct observation of previously hidden dynamic processes of life, allowing profound insights into mechanisms of health and disease. However, imaging of live samples is fundamentally limited by the toxicity of the illuminating light and images are often acquired using low light conditions. As a consequence, images can become very noisy which severely complicates their interpretation. In recent years, deep learning (DL) has emerged as a very successful approach to remove this noise while retaining the useful signal. Unlike classical algorithms which use well-defined mathematical functions to remove noise, DL methods learn to denoise from example data, providing a powerful content-aware approach. In this review, we first describe the different types of noise that typically corrupt fluorescence microscopy images and introduce the denoising task. We then present the main DL-based denoising methods and their relative advantages and disadvantages. We aim to provide insights into how DL-based denoising methods operate and help users choose the most appropriate tools for their applications.

Democratising deep learning for microscopy with ZeroCostDL4Mic.

Deep Learning (DL) methods are powerful analytical tools for microscopy and can outperform conventional image processing pipelines. Despite the enthusiasm and innovations fuelled by DL technology, the need to access powerful and compatible resources to train DL networks leads to an accessibility barrier that novice users often find difficult to overcome. Here, we present ZeroCostDL4Mic, an entry-level platform simplifying DL access by leveraging the free, cloud-based computational resources of Google Colab. ZeroCostDL4Mic allows researchers with no coding expertise to train and apply key DL networks to perform tasks including segmentation (using U-Net and StarDist), object detection (using YOLOv2), denoising (using CARE and Noise2Void), super-resolution microscopy (using Deep-STORM), and image-to-image translation (using Label-free prediction - fnet, pix2pix and CycleGAN). Importantly, we provide suitable quantitative tools for each network to evaluate model performance, allowing model optimisation. We demonstrate the application of the platform to study multiple biological processes.

Content-aware image restoration for electron microscopy.

Multiple approaches to use deep neural networks for image restoration have recently been proposed. Training such networks requires well registered pairs of high and low-quality images. While this is easily achievable for many imaging modalities, e.g., fluorescence light microscopy, for others it is not. Here we summarize on a number of recent developments in the fast-paced field of Content-Aware Image Restoration (CARE), in particular, and the associated area of neural network training, more in general. We then give specific examples how electron microscopy data can benefit from these new technologies.

Paired arrangement of kinetochores together with microtubule pivoting and dynamics drive kinetochore capture in meiosis I.

Kinetochores are protein complexes on the chromosomes, whose function as linkers between spindle microtubules and chromosomes is crucial for proper cell division. The mechanisms that facilitate kinetochore capture by microtubules are still unclear. In the present study, we combine experiments and theory to explore the mechanisms of kinetochore capture at the onset of meiosis I in fission yeast. We show that kinetochores on homologous chromosomes move together, microtubules are dynamic and pivot around the spindle pole, and the average capture time is 3-4 minutes. Our theory describes paired kinetochores on homologous chromosomes as a single object, as well as angular movement of microtubules and their dynamics. For the experimentally measured parameters, the model reproduces the measured capture kinetics and shows that the paired configuration of kinetochores accelerates capture, whereas microtubule pivoting and dynamics have a smaller contribution. Kinetochore pairing may be a general feature that increases capture efficiency in meiotic cells.

Kinesin-8 motors improve nuclear centering by promoting microtubule catastrophe.

In fission yeast, microtubules push against the cell edge, thereby positioning the nucleus in the cell center. Kinesin-8 motors regulate microtubule catastrophe; however, their role in nuclear positioning is not known. Here we develop a physical model that describes how kinesin-8 motors affect nuclear centering by promoting a microtubule catastrophe. Our model predicts the improved centering of the nucleus in the presence of motors, which we confirmed experimentally in living cells. The model also predicts a characteristic time for the recentering of a displaced nucleus, which is supported by our experiments where we displaced the nucleus using optical tweezers.

A divide and conquer strategy for the maximum likelihood localization of low intensity objects.

In cell biology and other fields the automatic accurate localization of sub-resolution objects in images is an important tool. The signal is often corrupted by multiple forms of noise, including excess noise resulting from the amplification by an electron multiplying charge-coupled device (EMCCD). Here we present our novel Nested Maximum Likelihood Algorithm (NMLA), which solves the problem of localizing multiple overlapping emitters in a setting affected by excess noise, by repeatedly solving the task of independent localization for single emitters in an excess noise-free system. NMLA dramatically improves scalability and robustness, when compared to a general purpose optimization technique. Our method was successfully applied for in vivo localization of fluorescent proteins.

Dynein motion switches from diffusive to directed upon cortical anchoring.

Cytoplasmic dynein is a motor protein that exerts force on microtubules. To generate force for the movement of large organelles, dynein needs to be anchored, with the anchoring sites being typically located at the cell cortex. However, the mechanism by which dyneins target sites where they can generate large collective forces is unknown. Here, we directly observe single dyneins during meiotic nuclear oscillations in fission yeast and identify the steps of the dynein binding process: from the cytoplasm to the microtubule and from the microtubule to cortical anchors. We observed that dyneins on the microtubule move either in a diffusive or directed manner, with the switch from diffusion to directed movement occurring upon binding of dynein to cortical anchors. This dual behavior of dynein on the microtubule, together with the two steps of binding, enables dyneins to self-organize into a spatial pattern needed for them to generate large collective forces.

Pivoting of microtubules around the spindle pole accelerates kinetochore capture.

During cell division, spindle microtubules attach to chromosomes through kinetochores, protein complexes on the chromosome. The central question is how microtubules find kinetochores. According to the pioneering idea termed search-and-capture, numerous microtubules grow from a centrosome in all directions and by chance capture kinetochores. The efficiency of search-and-capture can be improved by a bias in microtubule growth towards the kinetochores, by nucleation of microtubules at the kinetochores and at spindle microtubules, by kinetochore movement, or by a combination of these processes. Here we show in fission yeast that kinetochores are captured by microtubules pivoting around the spindle pole, instead of growing towards the kinetochores. This pivoting motion of microtubules is random and independent of ATP-driven motor activity. By introducing a theoretical model, we show that the measured random movement of microtubules and kinetochores is sufficient to explain the process of kinetochore capture. Our theory predicts that the speed of capture depends mainly on how fast microtubules pivot, which was confirmed experimentally by speeding up and slowing down microtubule pivoting. Thus, pivoting motion allows microtubules to explore space laterally, as they search for targets such as kinetochores.