![]() ![]() Touching and overlapping worms complicate feature extraction from individual animals. elegans worms cultured in 384-well plates imaged by bright field microscopy (kindly provided by collaborator Frederick M. Thin structures and low signal to noise make image segmentation and neuron tracing difficult, particularly when cells are more crowded than in this simplified example. (A) Projection of a 3D image of a fluorescently stained neuron (kindly provided by collaborator Mehmet Fatih Yanik). The system should also be useful to assess a potential drug’s human liver toxicity in order to prevent clinical trial failures. This enables large-scale experiments to identify chemicals that promote liver regeneration. If the fibroblasts are derived from mouse cells, the nuclei of the two cell types are distinctive enough to be distinguished by supervised machine learning. For example, human hepatocytes proliferate and maintain their native liver-specific functions much better when grown in the presence of fibroblasts. Developing computational approaches to analyze the complex images resulting from mixtures of two visually distinctive cell types is challenging but worthwhile. This type of co-culturing is also required for proliferation of some cell types. Biologists studying many different biological processes and diseases are increasingly making the extra effort to preserve natural cell–cell interactions by growing mixtures of physiologically relevant cell types together. MIXTURES OF CELL TYPES High-throughput experiments are inherently artificial in that they usually involve cells grown out of their native environment, typically in plastic multi-well plates. This is an invited paper in the special session on "Current challenges in image analysis for high-throughput microscopy."ģ. I highlight ongoing research areas of my group, the Imaging Platform of the Broad Institute of Harvard and MIT, where we focus on quantifying and mining the rich information present in high-throughput images (100,000–1,000,000 images per experiment) probing a variety of biological processes and diseases of interest. Here, some major challenges in this field are reviewed as part of an ISBI special session drawing attention to this growing area in biomedical imaging. Image processing algorithms and machine learning tools have been successfully employed to score increasingly complex phenotypes over the past decade. When each sample is imaged by microscopy, extracting the relevant, quantitative information from each image in an automated fashion becomes the main challenge. The goal is to “screen” the samples to identify those with desired effects. Each sample tests the effects of a particular gene or potential drug on a disease-relevant biological system, such as cells or small organisms. INTRODUCTION Due to advancements in robotic systems, biologists in pharmaceutical companies and academic screening centers are now able to efficiently create hundreds of thousands of biological samples in parallel. Index Terms- high-throughput, screening, fluorescence microscopy, co-cultures, C. Neuronal cell types are one of the “final frontiers” of two-dimensional mammalian cell image analysis. Furthermore, important information about neuron connectivity can only be gained by three-dimensional imaging, making experiments involving neurons a computational challenge in many respects. ![]() Foreground–background segmentation alone can be challenging, and tracing individual neurites that cross or are entangled is even more difficult. The state of the art is often to fall back on interactive guidance from the user, but this is infeasible for highthroughput experiments. Because the thin neurites that protrude from the cell bodies are often very weakly stained, automated algorithms often fail to accurately trace each neurite unless sample preparation and imaging conditions are optimal. One major exception is neuronal cell types (Figure 1A). This is true even for complex phenotypes, where machine learning has become indispensable. ![]() Scoring samples for a specific phenotypic change has become fairly routine in most mammalian cell types as well as similar-looking non-mammalian cells. The challenges include segmenting neurons, co-cultures of different cell types, and whole organisms segmenting and tracking cells in time-lapse images quantifying complex phenotypic changes and discovering biologically relevant subpopulations of cells.* In this paper, I describe some of these challenges, particularly those that are the subject of ongoing research in my laboratory. Major challenges remain in the extraction of rich information from high-throughput microscopy experiments. ![]() Carpenter Imaging Platform, Broad Institute of Harvard and MIT ABSTRACT EXTRACTING BIOMEDICALLY IMPORTANT INFORMATION FROM LARGE, AUTOMATED IMAGING EXPERIMENTS Anne E. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |