Posted on Thursday, June 28, 2012
Image-based research and development aim at exploiting the wealth of information contained in biomedical image data. While recent progress has dramatically facilitated the generation of large amounts of images, both the interpretation of the images and of the resulting data poses challenges. Traditionally, these challenges have been addressed separately with automated image analysis and data mining, respectively. Automated image analysis replaces or complements subjective visual assessment with consistent, quantitative readouts. The resulting numerical data is subsequently evaluated statistically to identify new hypotheses and verify conclusions.
At Definiens, we embrace this approach but challenge its sequential order. Drawing the right conclusions from images in research and development requires attention to subtle details, overview of very large data sets and, importantly, seamless transitions between the two. While image analysis can certainly capture subtle details and data mining provides an overview, often enough only data mining reveals the details worth paying attention to.
Definiens Image Miner™ supports seamless transitions between image analysis and data mining by providing interactive links between original images, image annotations and statistics. Novel hypotheses can thus be related to the image data they emerge from with no more than a few mouse-clicks. A product demo will emphasize that the resulting straightforward navigation through complex data sets substantially facilitates data review, quality control and validation of assays throughout the life sciences to reduce development costs and study times. The correlation of image analysis results with data from external sources provides a basis for the identification of novel diagnostics and biomarker tests using machine learning techniques as will be discussed using an image-based Her2 biomarker analysis as an example.
Webinar title: Integrating Data Mining with Image Analysis: Definiens Image Miner™ 2
Date: July 11, 2012
Time: 8am PDT / 11am EDT / 5pm CEST