Medical Imaging
Medical imaging is a crucial aspect of clinical care today. In every hospital or imaging center, a huge amount of data is generated on a daily basis from various modern imaging technologies. Further, the frequent use of multimodality imaging adds to the amount of data being acquired.
One of Numeric Insight’s noted strengths – Machine Learning – plays a vital role in the medical imaging field. Biological objects such as lesions do not come in standard shapes, nor can they be represented by equations. This inherent variability calls for a system that can “learn from examples” in order to perform common tasks necessary for diagnosis. Also, since medical images rarely yield clinical value when taken in isolation, integration with other biomedical signals and non-image information resources is needed.
New machine-learning algorithms are always being sought in the medical imaging field either to better solve diagnostically challenging problems or because new modalities and methodologies are constantly being introduced. As an example, consider molecular imaging which allows doctors to distinguish between normal and pathological biological processes by seeing the interactions of signaling molecules with sites on cell membranes and extracellular matrix.
Numeric Insight makes available to the medical imaging community its extensive experience in statistical pattern recognition, image segmentation and image registration to prototype new algorithms quickly and efficiently; and deploy tools as standalone or web-based applications.
Astronomy
Numeric Insight develops image processing tools to refine astronomy images as well as mathematical methods to manipulate models used in astronomy.
Today’s earth-based and space-based advanced imaging telescopes operate at the limits of their imaging capabilities to seek objects several orders to magnitude below the illumination level of the night sky. The raw data from these instruments are extremely far removed from the strikingly clear and colorful images that often grace the covers of popular publications. In fact, the instruments themselves introduce distortions that mask the true data.
Numeric Insight brings powerful tools and techniques to clean and process raw data extracting important information such as spectral signature which indicates the composition of the stellar atmosphere. Radio astronomy provides numerous opportunities for us to offer our capabilities in signal detection, deconvolution, parameter estimation and statistical analysis.
Audio Processing
The strengths of Numeric Insight in mathematical techniques to model patterns and structures in audio signals have been used to mitigate the effects of noise, distortion and replacement of missing samples.
Our capabilities include theory driven adaptive signal processing and optimal filter design, as well as practical considerations for performance within hardware boundaries such as available memory and processing resources and power.
Audio restoration, speech enhancement and noise reduction technologies play an increasing and important role in videoconferencing, hearing aids and acoustic communication systems. The routine use of hands-free telephony within cars brings with it a need to separate speech signals from unwanted background noise.
Automated Inspection
Instruments powered by computational methods developed by Numeric Insight have been used to perform accurate and efficient automated inspection. These are critical to applications seeking to find an optimal operating point driven by regulatory, quality, competitive and cost factors.
Companies that seek to gain by leveraging the economics of mass production place great importance on having better quality inspection backed by statistical data.
Our broad experience with image processing, image registration, blob analysis, object counting and sizing and pattern matching & recognition enables us to quickly prototype, test and build applications that work under non-ideal conditions such as poor camera resolution, changing camera perspectives and varying lighting intensity, color, or polarization.
Our applications have been used for PCB inspection, detecting vial cracks, inspecting surface quality and monitoring fill levels.
Bioinformatics
Numeric Insight provides answers to an urgent need for analysis and interpretation of massive amounts of data generated by bioinformatics systems. Commercial sequencers spew out billions of base pairs of sequence data points per shift, accumulating terabytes of unstructured textual and image information. Submerged in this ocean of data is vital information on important proteins that control essential parameters of the disease state being studied.
The tools we develop assist in fields such as sequence alignment, drug design and discovery, protein structure alignment, and prediction of gene expression. We excel at extracting distinguishing information from bioinformatics data obtained from normal and diseased cellular structures. We develop intuitive visualizations to show relationships among members of large data sets. This is challenging given that the data is often extremely noisy and will need custom preprocessing, especially with high-volume gene expression data.
We foster insight into clinically important biological processes by building algorithms to predict protein structure and/or function, or by building a family tree of related sequences. We create stand-alone, web-independent applications for biotechnology companies leading to stable, documented and supportable systems.
Biometrics
Numeric Insight’s strengths in pattern recognition, image processing and mathematical modeling are an asset to developers of biometric authentication systems. The better these systems are, the stronger is the level of trust placed by society in all manner of transactions needed in today’s complex world.
Biometric systems must be able to quickly access a sensor and read data on an individual’s unique pattern from fingerprint, iris, or even DNA. Algorithms then seek a match for the pattern amongst huge collections of stored patterns. This is rarely an easy task to accomplish reliably. For example, fingerprints lifted off a surface in a crime scene, the “latent fingerprints,” might have a complex background, smeared out ridge structure and overlapping impressions. A robust preprocessing algorithm can recover a sharp ink-like print using prior knowledge of ridge structures.
Creating algorithms that intelligently use more available information always leads to more accurate and robust systems. In fact, multimodality biometric systems that combine fingerprint, iris, DNA, head shape, voice spectrum, body language and heartbeat have shown the ability to better systems that rely on a single feature alone.
Digital Pathology
Numeric Insight offers opportunities for quantitative analysis of digitized pathology images through image processing and pattern recognition.
We use our expertise in segmentation and classification to build applications for pathology tasks such as quantification of antibody staining, recognition and classification of cells and multicellular objects.
Common, yet difficult, problems in histology image analysis are segmentations of cell nuclei, cell membranes and clinically significant structures within the cell. Another frequently encountered need is to align adjacent, perhaps differently stained, tissue image sections to each other. We support efforts to analyze pathology images to quantitatively characterize disease classification. This promotes the development of preventive and treatment plans customized to individual patients – personalized medicine.
Ocean Science
Numeric Insight supports data processing to provide critical information for naval applications and research on physical oceanography, climate change, biological stability and environmental stewardship.
The routine use of autonomous instruments enables abundant data harvesting at reduced cost. These instruments can continuously supply data on physical parameters and even biology. The data, of course, varies over space and time both microscopically and macroscopically. The availability of this data provides an opportunity to build models and study phenomena never attempted before.
The Arctic Ocean is an important piece of the climate puzzle, yet awaits remote data acquisition and processing which are impractical by traditional observing methods.
Our expertise in building numerically intensive applications and intuitive visualization tools enable researchers to identify, investigate and resolve important questions and advance our understanding of the ocean.