June 14, 2006 | DNA microarrays have revolutionized gene expression and genomic analysis by increasing throughput and allowing massively parallel analyses and comparisons. While this gene expression data is of value, expression patterns identified in this way must eventually be confirmed at the protein level in cells and tissues — whether one is doing drug discovery research or evaluating clinical specimens. Pathologists have traditionally devised numerous “semi-quantitative” grading systems for images created from slides that can be used on minimally automated imaging systems. The methods involve conventional microscopy and immunohistochemistry and are laborious, costly, and highly subjective.
More recently, the development of tissue microarrays (TMAs) has enabled a more quantitative and automated analysis. The challenge for automated analysis protocols is selecting a region of interest and normalizing expression signals so that they can be compared within and between data sets. Differences in subcellular localization add another dimension to the challenge for automated analysis. However, as the information capacity of computers has increased, automated quantitative morphometric analysis has become possible, and various companies have developed systems that are now on the market.
DMetrix provides an ultra-rapid array-microscope whole glass slide digital imaging system. Developed in collaboration with pathologists and digital imaging experts, the DMetrix DX-40 scanner can image 40 slides in an hour. The scanner takes only one ultra-high resolution image of the specimen(s) on the microscope slide to produce up to 12 gigabytes of image data. The image can serve immediately as input to a segmentation or morphometric-analysis algorithm.
| ||SCAN MULTIPLICATION: The |
DMetrix DX-40 scanner images
40 slides in an hour.
Technological advances in automated microscopes, digital image acquisition, and high-throughput screening techniques have also led to the need for more sophisticated software tools. DMetrix has teamed up with BioImagene, a provider of image management and analysis platforms, to integrate DMetrix’s DX-40 scanner with BioImagene’s Scientific Image Management System.
DMetrix and BioImagene announced the first implementation of their collaboration in partnership with IBM, for the Arizona Cancer Center, in March. The “Arizona” system accommodates up to 33 terabytes of image data and manages more than 100,000 high-resolution images. Images are produced at high throughput and resolution with DMetrix’s DX-40 scanner. Users access image data and metadata through a secure Web site. The Arizona system accommodates remote consultations for cancer prevention projects and sophisticated 3-D image analyses in drug discovery. It uses IBM eServer xSeries systems to operate all software and hardware components. IBM Tivoli Storage Manager automates data backup and archiving. BioImagine provides TissueMine software for image analysis and management, which is optimized for acquisition, processing, and analysis of images from tissue specimens. It is designed for identification and validation of targets in a variety of tissue types.
Typical drug safety trials also involve the screening of thousands of tissue slides — in this case for indications of toxicity. Icoria (formerly TissueInformatics, and now a part of Clinical Data) has developed automated pathology software called SlideScreen that uses machine vision techniques to automate structure and pattern recognition and the quantification of tissue staining. SlideScreen software screens and classifies tissues into groups of normal and non-normal with low false-negative and false-positive rates. Results can be automatically tabulated, graphically displayed, and made available for correlation with proteomic and genomic data sets. Icoria has also developed an automated software package for evaluation of the quality of histological slides — an important potential bottleneck for automated clinical pathology.
Histometrix’s AQUA software and microscope-based imaging platform use molecular co-localization rather than feature extraction to automate screening. TMAs are stained with fluorescently labeled antibodies, and the co-localization of the fluorescence signals within pixel-based subcellular compartments is measured. The process can determine whether a particular protein co-localizes with a membrane, nuclear, or other localized reference standard. Histometrix says the accuracy is comparable to an ELISA, allowing it to see relationships undetected by conventional pathologist-based analysis.
E-mail Robert M. Frederickson at email@example.com.