Menu [toggle]

Archived Seminars

Backlinks Print


May 15th
Computational Imaging in Cancer
Dr Julia Schnabel PhD, Chair in Computational Imaging, King's College, London

April 10th
Large-Scale Clinical MRI & Associated Technological Development.
Dr. Jacob Levman PhD, Canada Research Chair in Bioinformatics in the Mathematics, Statistics and Computer Science Department at St. Francis Xavier, NS.

Abstract: Boston Children's Hospital's (BCH) commitment to excellence in clinical research is exemplary. In 2007, Boston Children's Hospital replaced their entire clinical MRI suite with 3 Tesla Siemens systems providing abnormally high quality and consistent imaging in a clinical environment. Recognizing the potential of many MRI pulse sequences that would be relegated to pure research in other centres, BCH's clinical imaging department elected to include a variety of advanced MRI pulse sequences as part of routine clinical imaging (including high angular resolution diffusion imaging, resting state functional MRI and more). BCH also developed an interface to the hospital's clinical imaging database that provides access to all hospital affiliated researchers. This approach to clinical imaging research has supported a large‐scale analysis of clinical MRI data focused on healthy brain development (1,069 patients) as well as the world's largest autism MRI study (2,250 examinations) both of which will be featured in this presentation. Related technologies have also been developed to support ongoing neuroscience research based on the enormous amount of high quality clinical data at BCH. This includes software to provide tractography measurements distributed regionally across a patient's brain, hemispheric asymmetry analysis, novel machine learning formulations and regional brain maturation assessment on a per subject basis. Future work will continue large‐scale neuroscience analyses of clinical MRI data focused on the many neurodevelopmental disorders imaged with MRI at BCH (multiple sclerosis, cerebral palsy, ADHD, tuberous sclerosis complex, neurofibromatosis, migraine, language disorders and much more).


June 6th
Tumor Classification using Context-Aware? Features
Shazia Akbar PhD., Post-doctoral research fellow, New York University School of Medicine
Abstract: Identifying tumor automatically in histopathology images is a complex problem due to highly variable data and presence of complex structures and patterns. And yet, it is an important problem which can make a significant impact in digital pathology from automated tumor grading to big data analysis in clinical trials.
In this presentation I will describe a method to classify tumor using context features extracted from underlying image data and posterior probabilities. This method, "spin-context", iteratively fuses context with machine learning outputs to improve tumor classification performance. I'll show you some examples of spin-context when applied to a dataset of breast tissue microarray spots stained with estrogen receptor. Results show that with each iteration, context improves classification accuracy until convergence.

May 16th
Radiomics for the early prediction of recurrence after stereotactic ablative radiotherapy for lung cancer
Sarah Mattonen, PhD candidate in the Department of Medical Biophysics, Western University, London
Abstract: Stereotactic ablative radiotherapy (SABR) is a guideline-specified treatment option for early-stage lung cancer. However, significant post-treatment fibrosis can occur and confound the detection of local recurrence. The goal of this study was to assess physician ability to detect timely local recurrence, and compare physician performance with a radiomics tool. At 3 months post-SABR, radiomics could detect early changes associated with local recurrence that are not typically considered by physicians. This decision support system could potentially allow for early salvage therapy of patients with local recurrence following SABR.

April 20th
Cortical Perfusion: A Marker of Disease Severity in Multiple Sclerosis
Parsa Hojjat, PhD
Abstract: Multiple sclerosis (MS) is the most frequent cause of non-traumatic neurological disability in young and middle-age adults and the most common inflammatory demyelinating disease of the central nervous system. In total, 40NaV of MS patients experience cognitive dysfunction. Cortical involvement is reported in 59NaV of MS cases significantly contributing to the progression of both physical and cognitive disability, impairment. However cortical lesions are challenging to image clinically motivating surrogate techniques to assess cortical dysfunction in MS. This talk seeks to examine the utility of perfusion imaging in assessing cortical abnormality in relation to physical and cognitive impairment in MS.

Feb 22nd
Quantitative Perfusion and Permeability Biomarkers in Brain Cancer
Armin Eilagh PhD., postdoctoral research associate, Sunnybrook Research Institute
Abstract: Dynamic contrast enhanced (DCE) perfusion and permeability imaging, using computed tomography (CT) and magnetic resonance (MR) systems, is an important technique for assessing the vascular supply and hemodynamics of healthy brain parenchyma and tumours. These techniques can measure blood flow, blood volume, and blood-brain barrier permeability-surface area product and, thus, may provide information complementary to clinical and pathological assessments. They have been used as biomarkers to enhance the treatment planning process, to optimize treatment decision-making, and to enable monitoring of the treatment non-invasively. In this review, the principles of MR and CT DCE perfusion and permeability imaging are described (with an emphasis on their commonalities), and potential value of these techniques for differentiating high-grade gliomas from other brain lesions, distinguishing true progression from post-treatment effects, and predicting survival after radiotherapy, chemotherapy and anti-angiogenic treatments are presented.


December 10th
General Purpose GPU Computing and its Applications to Medical Image Processing
Robert Xu, PhD candidate, MBP, University of Toronto
Abstract: General-Purpose? computing on Graphics Processing Units (GPGPU) is the usage of GPU for tasks traditionally handled by the CPU. GPGPU has been able to accelerate non-graphical applications due to its natural parallel throughput architecture. In this talk, I will first briefly describe the background and fundamental ideas behind GPGPU. Then, I will present preliminary results pertaining to the use of GPU in medical image processing. Specifically, I will discuss the application of image registration, and how it can be used to perform motion correction in the context of image-guided interventions.

November 27th
Automated monitoring of activities of daily living to assess cognitive status
Babak Taati, PhD., Scientist, Toronto Rehabilitation Institute & Assistant Professor, Department of Computer Science at University of Toronto.
Abstract: Computer vision systems can play a role in providing care to individuals living with physical or cognitive disability. In this talk, I will first briefly review vision-based systems to provide assistance to older adults with dementia and to assist with usability studies for this population. I will then present preliminary results on assessing the cognitive status of older adults by way of monitoring common activities of daily living. Early identification of dementia can potentially lead to improved quality of life both for older adults with dementia and their family and caregivers who can better plan informal/formal care in advance.

October 28th
Segmentation of breast images: a work in progress
Anne Martel' PhD, Senior Scientist, SRI
Abstract: Image segmentation is often an essential pre-processing step in medical image analysis. Although the exact methods used vary depending on imaging modality and body site, there are some general approaches that have been shown to be very versatile across a wide range of tasks. Earlier methods concentrated on image features alone and then additional constraints were added to constrain solutions to conform to more realistic shapes. Atlas based methods became much more widely used once robust image registration methods became available and more recently the use of machine learning approaches is gaining in popularity. In this talk I will give a brief review of some of the main medical image segmentation methods available and then I will describe how (and why) our approach to breast segmentation in MR imaging has evolved over time.

June 24th
CT in Radiation Therapy.
William Song PhD, DABR, Head, Department of Medical Physics, Sunnybrook Health Sciences Centre

May 27th
Inverse Problems in Image Processing
Mehran Ebrahimi, PhD,'' Assistant Professor in the Faculty of Science at the University of Ontario Institute of Technology (UOIT).
Abstract: In many practical problems in the field of applied sciences, the features of most interest cannot be observed directly, but have to be inferred from other, observable quantities. The problem of solving an unknown object from the observed quantities is called an inverse problem. Many classical problems, including image reconstruction from samples, denoising, deblurring, segmentation, and registration, can be modelled as inverse problems.
Generally, many real-world inverse problems are ill-posed, mainly due to the lack of existence of a unique solution. The procedure of providing acceptable unique solutions to such problems is known as regularization. Indeed, much of the progress in image processing in the past few decades has been due to advances in the formulation and practice of regularization. This, coupled with progress in the areas of optimization and numerical analysis, has yielded much improvement in computational methods of solving inverse imaging problems.
In this talk, we will review general theoretical and computational aspects of inverse theory. Furthermore, we will revisit a number of inverse problems including image registration and present some recent research ideas.

April 22nd
Quantitative Ultrasound Monitoring of Locally Advanced Breast Cancer
Hadi Tadayyon, PhD candidate in the Department of Medical Biophysics, University of Toronto.
Abstract: Conventionally, the response of solid tumours to anti-cancer therapy is assessed using the RECIST guideline, which is based on gross size reduction. However, clinically significant tumor shrinkage does not occur until several weeks to a few months into chemotherapy treatment. We have proposed a novel non-invasive imaging method for early monitoring of locally advanced breast cancer patients using quantitative ultrasound spectral analysis techniques. We have demonstrated that patients’ tumor responses can be predicted as early as 4 weeks into treatment with an accuracy of 80%. This technique is relatively low-cost and can be extended to other cancer sites such as prostate and liver tumours.

March 18th
Quantitative Measurement of Masking in Mammography
James Mainprize, PhD Research Associate in Physical Sciences Platform, Sunnybrook Research Institute, Toronto.
Abstract: Mortality reductions as high as 40% are attributable to early detection by screening mammography. However, mammography has reduced accuracy for women with dense breasts. Lesions can be obscured or “masked” by surrounding or overlapping parenchymal tissue, or normal dense tissue can mimic the presence of a lesion. While lesion masking is prevalent in very dense breasts, it also occurs in non-dense or intermediate density breasts. It is suspected that local areas of density and the corresponding density patterns or texture reduce the conspicuity of a lesion. Ideally, if mammograms showing high masking potential could be identified, these women could be redirected to an alternative screening program. This may have a dramatic impact improving screening sensitivity and even reducing call-back rates and biopsies of particularly ‘difficult’ mammograms. To do this, an objective, quantitative measure of masking is required.
Using a mathematical observer model, the detectability (SNR) of simulated lesions can be calculated for small regions across the breast image. In areas of high masking, caused by low contrast or loss of conspicuity due to complex tissue backgrounds, the detectability will be low. Lesion contrast is calculated from signal differences between cancer tissue and the tissue transmission (density) map of the breast. Fitted power spectra are extracted from small ROIs (256×256) in raster fashion across the image of the breast, capturing both the quantum noise and the parenchymal tissue texture characteristics. From the contrast and power spectra, a simple “localized” observer model is used to estimate the detectability, dL (SNR) of a lesion centered in each ROI.
Strong correlation is seen between the average dL and volumetric breast density in n=138 mammograms. Correlation is seen between the tissue texture parameter—the spectral power-law exponent, β. In a small study of 24 cancers, there was a statistically significant (p=0.03) difference (38% change) in average dL between mammograms classified by an experienced radiologist as “easy” and “difficult” to interpret.
A measure of masking potential in mammograms has been developed and correlates with image characteristics that reduce lesion conspicuity and correlates well with a radiologist’s findings. Future work will involve using these masking measures as a means to separate useful mammograms from non-useful mammograms. Ultimately, this can be used in a stratified screening program that directs each woman to the most useful screening modality or screening frequency to maximize benefit.

Feb 18th
Image Segmentation Methods for Medical Imaging Applications
Eranga Ukwatta, PhD, Post-doctoral fellow, Sunnybrook Research Institute
Abstract: Medical image segmentation, a process of partitioning an image into multiple meaningful regions, is a critical step in image processing pipelines designed to extract quantitative information for diagnosis and treatment of diseases. However, image segmentation is prohibitive to be performed manually because manual delineations of 3D/4D structures require a significant amount of time and effort, involve considerable intra- and inter-observer variabilities which minimize the ability to detect small changes, and require a high level of expertise by the observer. In this presentation, I will describe the development of novel image segmentation methods based on 'convex max-flow formulations' that are specifically aimed at patient-specific analysis and modeling of cardiovascular structures and functions. I will also describe several multi-region-based segmentation methods that were developed for generating volumetric plaque burden measurements of the carotid and femoral arteries using non-invasive imaging methods, such as 3D ultrasound and magnetic resonance imaging. In addition to vascular imaging applications, image processing methods have a great potential to make an impact in the cardiac domain. My current research work aims to integrate image processing tools with computer modeling of the heart towards an overarching goal of developing a seamless pipeline for patient-specific modeling. Such models are built based on images of patient hearts, and would be utilized in clinics for numerous applications, such as the prediction of risk for arrhythmia. I will discuss about my recent work on segmentation and reconstruction of cardiac structures, which is a critical component in the process of building image-based patient-specific cardiac models. Finally, I will demonstrate several applications of these algorithms in other clinical domains including prostate and brain imaging.

Jan 26th
Texture Analysis and its Application in Brain MRI
Rouzbeh Maani, PhD, from Sunnybrook Research Institute
Abstract: Image texture is generally defined as intensity variations appearing in images. The high discriminative power of the texture features and the large number of applications that use these features have made texture analysis a major research trend for the last four decades. Some examples are in automatic product inspection, remote sensing, object recognition, image segmentation, and document processing. Recently, texture analysis has been increasingly noted for medical imaging applications including diagnosis and monitoring of different diseases such as epilepsy, Multiple Sclerosis (MS), and Alzheimer's disease. This talks gives a brief overview of texture analysis and shows its application in analysis of the brain MRIs. In particular, a novel voxel-based method for texture analysis of brain images is presented. The proposed method obviates the need for defining region of interest by providing a 3D statistical map comparing texture features on a voxel-by-voxel basis. The method provides a hypothesis-free analysis tool to study cerebral pathology in neurological diseases.


Dec 9th

Validation framework for quantitative perfusion imaging
Catherine Coolens, PhD, Assistant Professor at the University of Toronto, Department of Radiation Oncology and a Staff Medical Physicist in the Radiation Medicine Program at Princess Margaret Cancer Centre – University Health Network.
Abstract: Radiotherapy is the treatment of choice for a large number of cancer patients. Although modern techniques now tailor delivery based on a patient’s anatomy, individual variations in the behaviour of the tumor and surrounding healthy tissues are not fully considered. This motivates the goal of ‘personalized medicine’, which is potentially adapted to individual tumor characteristics. Dynamic Contrast-Enhanced? Computed Tomography (DCE-CT) is an imaging method that has been shown to have diagnostic and prognostic value in measuring tissue functionality. The underlying process is to measure the flow of intravenously administered CT contrast agent and the diffusion of these contrast molecules in the tissue-of-interest. Despite the relative successes of DCE-CT, the reproducibility and reliable quantification of these parameters has been difficult. Some of the main reasons for this are the simplistic algorithms; analysis methods and a lack of robustness in the DCE-CT measurements.
This seminar will discuss the development and validation of a 4D quantitative perfusion CT methodology with the long-term goal of maximizing its ability to characterize tumor morphology and physiology and to use this tool to probe tumor prognostic factors and response during therapy. This work ranges from technical scanner characterization and fluid dynamic modeling of contrast exchange to flow phantom development and clinical response assessment.

Oct 27th
Distributed Algorithms for Big Data Summarization
Ahmed Farahat, PhD, postdoctoral fellow at the University of Waterloo in Waterloo, Ontario, Canada.
Abstract: While deploying a data analytics solution, data scientists are typically overwhelmed by the massive amounts of unlabelled data, with no or little understanding of their structure. During this stage, there is an acute need to summarize this data by identifying the key features and data instances, as well as the natural groups that exist in the data. In this talk, I will present my recent research on developing distributed algorithms for two crucial tasks in big data summarization: subset selection and kernel clustering. First, I will present a novel algorithm for selecting a representative subset of data instances or features from large-scale data sets. The algorithm starts by learning a concise representation of the data using random projection, and then it greedily selects a few representatives such that reconstruction error of the concise representation is minimized. Experiments on benchmark data sets demonstrate the efficiency and effectiveness of the proposed approach in comparison to state-of-the-art methods for subset selection. Second, I will present scalable kernel k-means algorithms for clustering large and massively distributed data. These algorithms are based on the definition of a family of low-dimensional representations characterized by a set of computational and statistical properties. The proposed algorithms employ a unified parallelization strategy that first computes the corresponding representation of all data instances, and then clusters them in the new representation space. Empirical evaluations on medium and large scale data sets show that the proposed algorithms outperform other state-of-the-art methods for approximating kernel k-means.

Sept 10th
3D Fusion of Histology to Multi-Parametric? MRI for Prostate Cancer Imaging Evaluation and Lesion Targeted Treatment Planning
Aaron Ward, Ph.D. and Eli Gibson, Ph.D. Lawson Health Research Institute, Western University and Robarts Research Institute respectively
Abstract: Multi-parametric magnetic resonance imaging (mpMRI) of localized prostate cancer has the potential to support detection, staging and localization of tumours, as well as selection, delivery and monitoring of treatments. Evaluating the performance of mpMRI for prostate cancer imaging and delineation ideally includes comparison to an accurately registered reference standard, such as prostatectomy histology, for the locations of tumour boundaries on mpMRI. There are key gaps in knowledge regarding how to accurately register histologic reference standards to imaging, and consequently further gaps in knowledge regarding the suitability of mpMRI for tasks, such as tumour delineation, that require such reference standards for evaluation. This talk will cover our developed techniques for accurate co-registration of pre-surgery mpMRI with post-surgery digital histology images, which enables the fine-scale comparison of imaging with an accepted histologic reference standard for cancer location and Gleason grade. We will describe a method for quantifying the registration error, and will show how to operationalize this information using a novel statistical power calculation designed to incorporate registration error for planning of imaging validation studies. Finally, we will discuss our preliminary data resulting from the application of these techniques to the problem of determining the necessary focal therapy treatment margins to achieve desired histologic coverage of dominant intraprostatic lesions. This talk is intended both for engineering/physics researchers who are interested in applications of image registration, as well as for clinician scientists interested in targeted prostate interventions and pathology workflows enabling accurate imaging-histology correlation.

March 27th
Scalable Non-Linear? Dimensionality Reduction by Isometric Patch Alignment
Ali Ghodsi, PhD, Associate Professor in the Department of Statistics at the University of Waterloo.
Abstract: We propose a novel dimensionality reduction method that is scalable and has low computational cost. This method is inspired by two key observations: (i) the structure of reasonably large patches of high-dimensional data can be preserved as a whole, rather than divided into small neighborhoods; and (ii) attaching two neighboring patches will align them such that the overall rank does not increase. In the proposed approach, the data is divided into smaller clusters ( and can be distributed to different machines). Each cluster (on a different machine) is embedded into a low-dimensional patch and then all of the patches are rearranged such that their border points are matched. We show that the rearrangement can be computed by solving a relatively small semidefnite program. The embedding computed by this optimization is provably low-rank. The proposed method is stable, fast, and scalable; experimental results demonstrate its capability for dimensionality reduction data visualization, and even complex tasks such as protein structure determination. This is a joint work with Pooyan Khajehpour.


March 14th
ClearCanvas: An extensible imaging applications platform for clinicians and researchers.
Norman Young, CEO and Founder of ClearCanvas?
Abstract: ClearCanvas? is a multi-faceted medical imaging software that offers both clinicians and researchers standard imaging functionalities such as PACS connectivity and visualization (MPR). In addition to the management and viewing functionality, the platform is an extensible tool. The software has been specifically designed to provide a platform on which researchers and software developers can quickly add their own custom functionalities. ClearCanvas? enables clinicians, researchers, and medical imaging software developers to be a creator and partner by using the open source code to develop and add their own customized algorithms to the software and/or to be a user by incorporating a ClearCanvas? commercial product into the clinical or research setting for image management, access to images and workflow efficiencies.

Feb 14th
Developing image-guided robotics for breast intervention
Peter Bevan, PhD, McMaster? University
Abstract: This presentation will provide an overview and current status of development of an MRI-compatible automated robotic breast intervention system developed by the Centre for Surgical Invention and Innovation (CSii) in collaboration with MDA Robotics. This image-guided prone biopsy system (IPBS) completed prototype development and testing Dec 2011, and the final system will be ready for clinical studies in the summer of 2013.
Topics to be discussed:
- An overview of design and testing of the MRI-compatible robotic manipulator prototype
- Design considerations and implementation of the final MRI-compatible robotic breast intervention system
- Plans for multi-centre clinical study of the safety and effectiveness of the system
- Research and development activities with academic collaborators intended to enhance future capabilities of the system.
Anticipated Impact:
Magnetic resonance imaging (MRI) has been shown to detect some breast cancers that are undetectable to standard mammography. Due to the high sensitivity, it is recommended as as screening modality for high-risk patients. After suspicious tissue is detected by imaging, the next step is to perform a biopsy. The current MRI-guided breast biopsy procedure is time consuming and requires a skilled radiologist to perform effectively. IPBS will offer an accurate and straightforward biopsy procedure, with the opportunity for immediate treatment (ablation) of suspicious tissue. This will give the option to perform a more efficient "biopsy and ablation" procedure, which will optimize time in the MRI suite, shorten patient wait times between procedures, and reduce overall burden to the health care system.

January 10th
Lesion Explorer
Joel Ramirez email
Subcortical hyperintensities (SH) are a commonly observed phenomenon on MRI of the aging brain and are believed to indicate some form of underlying small vessel disease. The Lesion Explorer image processing pipeline yields regionalized volumetrics for intracranial brain tissue and various SH lesion subtypes. Presented by Dr. Joel Ramirez, a post-doctoral fellow working with Dr. Sandra Black at Sunnybrook Research Institute.
LesionExplorer-SMIAL_Jan<a href=tiki-editpage.php?page=LesionExplorer-SMIAL_Jan title=Create page: LesionExplorer-SMIAL_Jan class=wiki wikinew>?</a>.10.2013.pdf (2.99 MB)


Dec 14th
Supervised Dictionary Learning
Mehrdad Gangeh, University of Waterloo
Abstract: Dictionary learning and sparse representation are two related topics which lead to state of the art results in many computer vision and machine learning applications. In this talk, we first briefly provide an introduction to these topics. Then we propose a supervised dictionary learning (SDL) technique by incorporating information on class labels into the learning of dictionary. To this end, we propose to learn the dictionary in a space where the dependency between the signals and their corresponding labels is maximized. To maximize this dependency, the recently introduced Hilbert Schmidt independence criterion (HSIC) is used. One of the main advantages of this novel approach for SDL is that it can be easily kernelized by incorporating a kernel, particularly a data-derived kernel such as normalized compression distance, into the formulation. The learned dictionary is compact and the proposed approach is fast. We show that it outperforms other unsupervised and supervised dictionary learning approaches in the literature on real-world data.

9th November
IPython: tools for the entire lifecycle of research computing
Dr. Fernando Perez, Helen Wills Neuroscience Institute, University of California, Berkeley.
Abstract: The IPython project ( started as a better interactive Python interpreter in 2001, but over the last decade it has grown into a rich and powerful set of interlocking tools aimed at enabling an efficient, fluid and productive workflow in the typical use cases encountered by scientists in everyday research.
In this talk we will show how IPython supports all stages in the lifecycle of a scientific idea: individual exploration, collaborative development, large-scale production using parallel resources, publication and education. In particular, the IPython Notebook supports multiuser collaboration and allows scientists to share their work in an open document format that is a true "executable paper": notebooks can be version controlled, exported to HTML or PDF for publication, and used for teaching. We will demonstrate the key features of the system, including recent examples of scientific publications made with the notebook.

Oct 11th
Optimizing Pre-processing in fMRI: improving spatial reliability and behavioural correlations
Nathan Churchill, Baycrest
Abstract: BOLD fMRI is an invaluable tool for measuring the haemodynamic correlates of brain function, with applications in both the clinical and experimental domains. However, this technique remains limited by a relatively poor Contrast-to-Noise? Ratio, and noise sources that may exceed signal by an order of magnitude, including subject movement and physiology. These experimental noise sources are difficult to control, as they often have large temporal variance, they may be correlated with the BOLD response, and their effects vary across subjects, tasks and even sessions. It is thus difficult to perform accurate, reliable measurements using BOLD fMRI.
Although a variety of pre-processing techniques have been developed to correct for noise in fMRI, there is no literature consensus on a single, fixed set of “best” pre-processing steps. Indeed, there is evidence that the set of preprocessing choices (or “pipeline”) should be adapted on an individual-subject basis, in order to optimize signal detection. I will demonstrate how our framework, which uses data-driven measures of spatial reproducibility and prediction accuracy, can be used to identify the optimal preprocessing pipeline on an individual subject basis. Furthermore, I provide evidence that pipeline optimization significantly improves a number of relevant measures in fMRI, including (1) between-subject reliability of activations, (2) within subject test-retest reliability and (3) correlations with behavioural measures. These results indicate that typical fixed-preprocessing strategies may significantly limit the sensitivity of BOLD fMRI analyses.

19th July
Developing General Computational Frameworks for Image Registration
Dr. Nathan Cahill, Associate Professor, Center for Applied and Computational Mathematics in the School of Mathematical Sciences at the Rochester Institute of Technology.
Abstract: Many of the algorithmic developments in the field of image registration have emerged from work on specific applications, such as PET/CT/MR brain imaging, contrast enhanced MR breast imaging, 3D reconstruction from histological sections, etc. While algorithms that are tuned to specific applications can certainly exhibit good performance, it can be useful to draw ideas from these algorithms in order to formulate a general computational framework that could be applied in a wide variety of settings. In this talk, we show how general computational frameworks can be developed for image registration that can handle a wide variety of image/feature similarity measures, regularization techniques, boundary conditions, and constraints. We will use various application settings to illustrate the results, including brain images, breast images, and histological sections.

7th July
Mathematical image processing techniques: ideas, theory and applications
Prashant Athavale, Ph.D (Postdoctoral Fellow, Sunnybrook Research Institute)


Mar. 4
Inverse Problems in Image Processing
Mehran Ebrahimiemail
mehran-smial-talk.pdf (2.71 MB)

Mar. 26
Why Open Source Will Rule Scientific Computing
William Schroeder email (5.39 MB)

His series of blog posts (external link) on this topic provides more detail.


Feb 5th
Extracting information from contrast–enhanced medical images using data-driven methods
Anne Martel pdf (1.92 Mb)

Mar. 5th
An Introduction to Mutual Information Based Registration.
Perry Radau email
SMIAL-MutualInfo-March2009.pdf (2.48 MB)

May 7th
Automatic Contrast Enhancement and Segmentation of Cerebral White Matter Lesions in FLAIR MRI.
April Khademi
2009-05-07 - Pres - Contrast Enhancemnt of WML - SMIAL Pres.pdf (2.48 MB)

Nov. 5th
Open Science and the X Prize with Application to Cardiac Segmentation.
Perry Radau email
SMIAL Open Science and Cardiac Segmentation 2009Nov5.pdf (2.83 MB)

Created by: amartel Last Modification: Monday 09 of October, 2017 15:35:58 EDT by amartel

Upcoming events