Guest Editor's Introduction: Multimedia Signal Processing and Systems in Healthcare and Life Sciences
Issue No.04 - October-December (2007 vol.14)
Published by the IEEE Computer Society
Nevenka Dimitrova , Philips Research
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MMUL.2007.73
The guest editor discusses the articles contained in this special issue and explains how this special issue reflects trends, challenges, and opportunities in the field at large.
Signal processing as a traditional discipline offers powerful and well-studied tools for analyzing spatial and temporal data. Lately there's been great interest in the life sciences and healthcare communities to process, store, and analyze continuous signals from a variety of sources. Examples include modalities in imaging diagnostics (ultrasound, x-ray, computerized tomography, magnetic resonance imaging, and functional MRI), continuous physiological monitoring, audio and video monitoring, and high throughput measurements of the processes at a molecular level.
Each modality presents its own problems, and combined, analysis of all this data can enable fast discoveries in genomics, biology, and the medical field. This special issue has a wonderful blend of articles that explore the utility of multimedia signal processing in life sciences as well as in healthcare applications.
A closer look
On the life sciences side, at the molecular level, microarray data analysis is currently a hot topic in functional genomics and systems biology. The measurement of the activity of thousands of genes with only a few microarray samples makes it a feature-rich and case-poor type of problem and requires powerful computational tools. In "Learning Microarray Gene Expression Data by Hybrid Discriminant Analysis," Lu et al. have developed a new computational method of dimension reduction to classify microarray gene expression data. They present a method that combines the strengths of principal component analysis and linear discriminant analysis for learning microarray gene expression data.
Also in life sciences, to get a complete picture of the genotype–phenotype relationship in the context of the nervous system it's important to quantify animal behavior. Since high-throughput experimental set ups increasingly require automated acquisition and analysis of behavioral data, multimedia systems are becoming a vital part of the behavioral neuroscientist's tool kit. In "Characterizing Animal Behavior through Audio and Video Signal Processing," Valente et al. explore the characterization of animal behavior through the use of audio and video signal processing. This type of objective description is essential for properly interpreting results and ensuring reproducibility of experiments across laboratories.
The availability of multimedia processing has been precipitated in imaging applications in the clinical setting. In "Positioning Tasks in Multimodal Computer-Navigated Surgery," Brell et al. introduce positioning tasks in multimodal computer-navigated surgery as an approach for communicating information to a surgeon during an interventional procedure using digitally augmented visuals and haptics.
The novel aspect of this work is the manner in which the haptic communication channel is used in addition to the surgical tool while superimposing preoperative computed tomography and medical resonance imaging data. An optical tracking system monitors the position of the surgical tool and the surgeon's hand, relative to the patient's anatomy and relative to the preoperative plan. Tactile feedback is then provided to the surgeon (via a set of haptic actuators arranged on the back of the surgeon's hand/fingers), which direct hand motions that will correctly execute the surgical plan.
At a higher abstract level, there's a strong need to apply known multimedia modeling techniques to the field of biology and medicine. In "Knowledge Extraction for High-Throughput Biological Imaging," Ahmed et al. introduce a multilayered architecture and spatiotemporal model for the analysis of biological images. They apply their spatiotemporal modeling-based knowledge extraction for high throughput biological imaging. The model the authors proposenot only extracts low-level features from biological images but also the high-level information such as biological events from image sequences.
Some final thoughts
In summary, I'm happy to present this nice array of articles as one of the first introductions to what's possible at the confluence of multimedia signal processing and problem areas of the life sciences and healthcare. Through this research, maybe one day we can change the way we make new discoveries in life sciences, particularly where we need analysis and integration of large amounts of multimedia data. In healthcare, by leveraging diagnostics, we can foresee a new workflow that brings predictive and personalized medicine closer to reality.
Nevenka Dimitrova is a research fellow in the Department of Reliable Care Solutions at Philips Research, Bangalore, India. Her current research activities are in bioinformatics and biomarker discovery and how to enable decision support systems for personalized medicine. At the same time she has been a visiting scientist at Cold Spring Harbor Laboratory and at Columbia University. Dimitrova obtained her MS and PhD in computer science from Arizona State University. She has 43 issued patents and acted as general chair of the ACM Multimedia 2004 conference.