CBIIT Speaker Series Page

SYNOPSIS:

Genomic analysis of individual patients is now affordable and therapies targeted to specific molecular aberrations are being tested in clinical trials. However, even highly-specialized physicians at leading academic centers are not equipped to apply genomic information available in publicly available sources to clinical-decision-making concerning individual patients. Dr. Meric-Bernstam will describe informatics tools she and her team have developed to support personalized cancer treatment as "standard of care" rather than "one off" exceptions. These include: 1) a standardized approach for classification of variant actionability, 2) a database of therapeutic implications of common genomic aberrations using automated processing of publicly-available sources and 3) tools to summarize and present patient-specific annotations to clinicians. Efforts to implement these tools at a large comprehensive cancer center and beyond as well as approach to provider education in genomics and decision support tools will be reviewed.

Session details...

SYNOPSIS:

Over the last couple of decades, “challenges” have been successfully employed to spur scientific research, “leverage ingenuity” and foster the translation of scientific advances into more widespread use. The topics for the challenges have covered a large spectrum of critical issues from self-driving cars and robots for “dangerous, degraded, human-engineered environments” to topics in energy, education and human health.  

“Challenges” have also becoming increasingly important in the medical imaging research community. Such challenges have been an integral part of prestigious conferences such as MICCAI (Medical Image Computing and Computer Assisted Intervention) and International Symposium on Biomedical Imaging (ISBI) and are being planned at a number of other venues. The underlying rationale for these challenges is driven by the realization that every year we see the publication of numerous algorithms published in the scientific literature, yet a very small fraction are translated into clinical use. Challenges can be an effective means to comprehensively assess the performance of algorithms by comparing them on common, sufficiently large and diverse datasets using realistic tasks and valid evaluation metrics.

MedICI is an open-source project that is developing infrastructure and support to  host medical imaging challenges across radiology, digital pathology, and genomics. We will describe the architecture of the system including the integration of CodaLab, caMicrosocope, and ePAD. We will walk through the process of hosting and participating in challenges from the perspective of the organizer and participant, describe past and on-going challenges and share successes as well as lessons learned.

Session details...

SYNOPSIS:

caMicroscope and DataScope is one of the three Clinical and Translational Informatics Projects that were funded by NCI/NCIP. This project had two distinct, but integrated goals namely: a) caMicroscope — A digital pathology platform that supports visualization, annotation and analysis of digital pathology data; and b) DataScope — an interactive data integration, query and exploration system. In this talk I will be doing a deep dive into the capabilities of both these systems. caMicroscope provides the community with an open source solution that can visualize whole slide pathology images, create and display both human and machine generated annotations, and run analysis algorithms on the images. In this talk I will provide an overview of caMicroscope, summarize some of it’s deployments, and provide a roadmap for upcoming features. DataScope is part of the Integrative Query System and provides an interactive environment to integrate and explore disparate datasets. Providers can use DataScope to create rich exploration systems that end users can use to slice-dice the underlying datasets, in a highly declarative fashion, without any software development. The talk will touch upon some of its recent deployments and upcoming features.

Session details...

SYNOPSIS:

This talk will describe the challenges in the area of scientific workflows, including how they are used to advance science in a number of domains, and how state-of-the-art software systems, such as Pegasus, meet the application and computing infrastructure challenges.  Pegasus enables scientists to describe the workflows in an abstract, resource-independent way. That description includes the definition of the workflow steps and the data they take in and generate, but does not include low-level cyber-infrastructure information. Given the abstract workflow description and the information about the execution environment (composed of potentially distributed data sources and systems), a planner can map the computational tasks onto the available resources and plan the movement of data across distributed resources. The planning process also opens up opportunities for performance optimization and fault-tolerance. The talk will describe example applications, including LIGO, the gravitational-wave physics experiment that recently confirmed the existence of gravitational waves. The talk will touch upon the issues the applications face, and how Pegasus can help them execute in a number of different environments: campus clusters, distributed resources, and clouds.

Session details...