Image-based screening in biomedicine; current methods and future challenges
Principal investigator at the Imaging Platform of Harvard and MIT, Cambridge,
MA Associate Professor in Quantitative Microscopy at Uppsala University, Sweden
About the organizer
The use of image-based large scale high-throughput and high-content screens to test chemicals (drugs or toxicity assessment) and/or the effect of genetic perturbants (RNAi) is common practice in pharma-industry as well as academic research labs today, and when it comes to cell-based screens, there is a set of standard algorithms used in almost all commercial and open-source solutions. These include pre-processing (mainly illumination correction), image segmentation, and feature extraction. An important aspect prior to analysis is image quality control and methods for handling non-optimal data. These subjects will be covered in the first half of the tutorial. Next, developments and challenges in the field, such as HTS on model organisms (C. elegans and zebrafish), more complex cell-based assays (neurons, co-cultures, 3D), tissue analysis, and screens using electron microscopy will be discussed. Finally, open-source software (mainly CellProfiler and ImageJ/Fiji) for image-based screening will be presented, and approaches for contributing state-of-the-art methods, such as those presented at ISBI, to the screening community will be discussed. Collections of benchmarking datasets, encouraging algorithm developers to compare their methods in a reproducible way, will also be presented.
Co-presenter:
Ida-Maria Sintorn,
Associate Professor in Computerized Image Analysis at Swedish University of Agricultural Sciences
Ida-Maria Sintorn received her PhD in Computerized Image Analysis from the Swedish University of Agricultural Sciences in 2005, focusing on segmentation and shape description methods in 2D and 3D microscopy. She spent two years as an image analysis researcher in the Biotech Imaging group at the industrial research organization CSIRO in Sydney, developing methods for mainly 2D and 3D analysis of neurons and neuronal co-cultures. Since 2007 she heads the image analysis group at a small biotech company in Stockholm performing electron microscopy based virus and nanoparticle analyses, while also developing methods for automated electron microscopy and virus identification as Associate Professor at the Swedish University of Agricultural Sciences.
Assistant Professor, Department of Statistics, Department of Computer and
Mathematical Sciences, Department of Computer Science University of Toronto
About the organizer
Ruslan Salakhutdinov received his PhD in computer sciencefrom the University of Toronto in 2009. After spending twopost-doctoral years at the Massachusetts Institute of Technology Artificial Intelligence Lab, he joined the University of Toronto as an Assistant Professor in the Departments of Statistics and Computer Science.Dr. Salakhutdinov’s primary interests lie in artificial intelligence, machine learning,deep learning, and large-scale optimization.He is the recipient of the Sloan Research Fellowship,Early Researcher Award, Connaught New Researcher Award,and a Scholar of the Canadian Institute for Advanced Research.
Building intelligent systems that are capable of extracting high-level representations from high-dimensional sensory data lies at the core of solving many AI related tasks, including visual object or pattern recognition, speech perception, and language understanding. Theoretical and biological arguments strongly suggest that building such systems requires deep architectures that involve many layers of nonlinear processing.
Many existing learning algorithms use shallow architectures, including neural networks with only one hidden layer, support vector machines, kernel logistic regression, and many others. The internal representations learned by such systems are necessarily simple and are incapable of extracting some types of complex structure from high-dimensional input.
In the past few years, researchers across many different communities, from applied statistics to engineering, computer science and neuroscience, have proposed several deep (hierarchical) models that are capable of extracting useful, high-level structured representations. An important property of these models is that they can extract complex statistical dependencies from high-dimensional sensory input and efficiently learn high-level representations by re-using and combining intermediate concepts, allowing these models to generalize well across a wide variety of tasks.
The learned high-level representations have been shown to give state-of-the-art results in many challenging learning problems, where data patterns often exhibit a high degree of variations, and have been successfully applied in a wide variety of application domains, including visual object recognition, information retrieval, natural language processing, and speech perception. A few notable examples of such models include Deep Belief Networks, Deep Boltzmann Machines, Deep Autoencoders, and sparse coding-based methods.
The goal of the tutorial is to introduce the recent and exciting developments of various deep learning methods to the ISBI community. The core focus will be placed on algorithms that can learn multi-layer hierarchies of representations, emphasizing their applications in computer vision and biomedical image analysis.
Motion-Robust Super-resolution Magnetic Resonance Imaging
Director, Computational Radiology Laboratory,
Professor of Radiology, Harvard Medical School
Director of Research of the Department of Radiology at Children’s Hospital BostonAli Gholipour
Instructor of Radiology Harvard Medical School
Member of the Computational Radiology Laboratory at Children’s Hospital Boston.
Additional speakers: Benoit Scherrer, Onur Afacan
Website
http://crl.med.harvard.edu/research/isbi2013_tutorial/
Technical innovations overcoming the limitations of existing medical imaging technology will enable improved diagnosis, monitoring and therapeutic intervention assessment in medicine. Ultimately, it will offer better clinical care for patients. Magnetic resonance imaging (MRI) is a non-invasive imaging modality that generates a unique range of contrast to evaluate many organs, structures, and anomalies in vivo. The use of MRI, however, has been limited mainly by two factors: the relatively low spatial resolution achievable and its sensitivity to motion.
The sensitivity to motion makes it highly challenging to acquire good quality scans when imaging newborns, children and non-cooperative patients. In clinical practice, sedation and anesthesia can be used but lead to significantly increased risks, burden and costs. Poorly cooperative subjects for which there is no clear direct benefit justifying the sedation cannot generally be imaged. Novel developments in research are necessary to enable high quality scans in presence of motion.
The other major limitation of MRI is its spatial resolution which is often chosen relatively low in a compromise to achieve high signal-to-noise ratio (SNR). Acquisition of high spatial resolution MRI with high SNR requires that the patient or the organ remains completely still in the scanner for an extensive period of time. High resolution in-vivo MRI is strongly challenged by the susceptibility of MRI to motion. The problems of motion and resolution in MRI are inter-connected: motion robust imaging enables longer acquisitions to improve the resolution, while resolution enhancement techniques can be used to image at conventional resolution with faster scans, reducing the impact of motion.
Extensive research has been carried out for motion compensation and also for resolution enhancement in MRI; however, these issues have rarely been considered together. The recent literature suggests that images of moving subjects and moving organs may be obtained at high spatial resolution using integrated motion-robust super-resolution reconstruction. Motion-robust super-resolution MRI can reduce the risk and the cost of MRI by reducing the rate of sedation and repeated acquisitions and may significantly increase the capability and capacity of MRI in clinical evaluation as well as in research studies.
Statistics of Medical Imaging: Theory and Applications
Research Associate Professor
Dept. of Radiology, University of PittsburghPhysical principles and mathematical procedures of medical imaging techniques such as CT, MRI, PET, SPECT, and US, etc have been extensively studied during past decades. Various image processing and analysis methods such as Graph, classical snakes and active contour, Level set, active shape and appearance models, fuzzy connected object delineation, random field, etc have been applied in clinical medicine and basic research. With these thorough and in-depth research and development has come a realization of the need for a complete, statistical study of medical imaging. Statistical investigation into medical imaging not only provides a better understanding of the nature of the technology (analysis), but also leads to an improved design of the technology (synthesis).
This tutorial provides a theoretical framework of statistical study into medical imaging. It
- Describes physical principles and mathematical procedures of two medical imaging techniques: X-ray CT and MRI,
- Presents statistical properties of imaging data (measurements) at each stage in the imaging processes of X-ray CT and MRI,
- Demonstrates image reconstruction as a transform from a set of random variables (imaging data) to another set of random variables (image data),
- Presents statistical properties of image data (pixel intensities) at three levels: a single pixel, any two pixels, and a group of pixels (a region),
- Provides two stochastic models for X-ray CT and MR image in terms of their statistics and two model-based statistical image analysis methods,
- Evaluates statistical image analysis methods in terms of their detection, estimation, and classification performances,
- Indicates that X-ray CT, MRI, PET and SPECT belong to a category of imaging: the non-diffraction computed tomography.
Note: “This tutorial will be based on the book “Statistics of Medical Imaging” (CRC, 2012). Please contact the tutorial organizer for discounted advance purchase from CRC.”
