The Quantitative Microscopy Laboratory (QML) focuses on instrumentation, algorithms, and techniques for high-speed, fully automated optical microscopy with application to cell and tissue analysis. Previously labor-intensive processes requiring exhaustive analysis of large specimen areas are made easier thanks to high-performance autofocus, fully automated image segmentation, and precise image fluorometry. High priority research applications include detection of ultra-rare fetal nucleated red blood cells in the maternal bloodstream for cytogenetic testing and of micrometastatic breast cancer cells in peripheral blood for minimally invasive cancer screening, detection and purging of contaminant cells in cell culture mixtures, automated classification of cell populations for cancer screening, imaging and tracking of live cells correlated with cellular behavior and protein activation, and tissue microarray analysis.

   Biological specimens do not all lie in the same focal plane, complicating automated cytometry. One high-performance feature of our systems is on-the-fly autofocus. Autofocus involves analyzing frequency content from several optical sections to determine best focus. Optical sectioning can be done by sequentially moving the objective from one plane to another-a new design enables simultaneous acquisition of several optical sections using a system of prisms, mirrors, and CCD cameras. Autofocus can then keep the image plane sharply in focus during continuous stage travel, and additional cameras can collect fluorescent images in parallel. This continuous scanning system allows imaging 10-times faster than in a sequential-scanning cytometer, with dwell time and light sensitivity 10-fold greater.

   To enable run-time image segmentation of cells from background, a convolution filter was previously developed for 2D imaging, allowing contrast enhancement and automatic thresholding. The 2D filter was trained using "ideal" segmented image contours (based on actual cell images) using active contours-based software developed in our laboratory. The 2D contrast-enhancing technique is now extended to 3D, starting with fluorescent beads of known shapes and sizes to train the filter.

   Moreover, advances have been made in 3D cytometry. Three-dimensional specimen images are usually created from stacks of axially displaced 2D images. High axial resolution stacks have been obtained using the high-throughput Digital Micromirror Device (DMD) parallel laser scanning confocal microscope (P-LSCM). The DMD is an array of 16x16 μm² mirrors that serve as both point sources and detection pinholes. Using the DMD, light from multiple pinholes can be acquired simultaneously. Through parallel scanning, light collection efficiency and frame rate can be 100-fold higher than in a single laser scanning confocal microscope (S-LSCM). High sensitivity in the DMD system allows high-speed confocal imaging at intensities below the cellular fluorotoxicity threshold for living tissue.

   The fully automated scanning cytometer is currently involved in clinical trials for detecting ultra-rare fetal nucleated cells in maternal blood. Fetal cells exist in the maternal circulation at frequencies as low as 1:10⁵-1:10⁹ of maternal nucleated cells. The presence of fetal hemoglobin and a nucleus discriminate for these cells, and their fetal origin is confirmed by fluorescence in situ hybridization (FISH) on X and Y chromosomes in male pregnancies. Analysis of these ultra-rare cells could eliminate the need for invasive prenatal diagnostic techniques such as amniocentesis and chorionic villus sampling. Fetal cells have been detected with a false negative rate of 1 in 15 million and a false positive rate of 12 in 1 million as compared to rates on the order of 10⁴ per million for flow cytometry. This instrument is also being used to detect breast cancer cells in a peripheral blood model for diagnostic prescreening.

   Cell-by-cell classification is important for cytopathological decision-making. Automated cell analysis can be achieved, for example, by estimating the probability of occurrence of a certain class given particular features of a microscope specimen. A wide variety of features can be used: morphometric features (e.g. area, aspect ratio), photometric or fluorometric features (e.g. total optical density or integrated intensity), and texture (e.g. high-/medium-/low-density chromatin regions). In specific cell analysis applications, our instrument helps determine the significant feature set leading to the most accurate classification. Currently, cell classification is applied to detecting differences in cellular compartment patterns between cancer and normal cells.

   A prototype system for laser-ablation has been developed. A high energy, pulsed laser is directed to photoable all unwanted cells. One important clinical application is the purging of contaminating tumor cells in autologous bone marrow transplants. The system can also be used for optoporation. In this process, a lower power laser pulse transiently permeabilizes the cell membrane. Large protein molecules diffuse into cells without the cells dying. Optoporation could find use in selective drug delivery to target cells and for labeling enzyme pathways to elucidate gene/protein functions.

   Biologists have long studied cellular and molecular events in vitro and single cells in vivo. Direct observations of intra-cellular functions correlated to behavior are more realistic, but current fluorescence measurement techniques make acquiring statistically relevant data very laborious. The scanning cytometer serves as a perfect platform for live cell imaging and tracking. Repeated scanning of multiple fields of view enables automated time-lapse analyses of cellular migration and protein activity in hundreds to thousands of cells. Cellular biosensors (such as the Rho-family of proteins) that demonstrate local changes in fluorescence upon protein activation can easily be measured by time-lapse microscopy and correlated to cellular behaviors.

   Another area of active research involves instrumentation and numerical analysis methods for Tissue Microarrays (TMAs). TMAs are a new technology for rapid analysis and discovery of protein markers, especially for cancer detection. Currently in place are methods for acquiring high-resolution images of tissue samples and identifying normal vs. cancerous regions. Research challenges include improving high-throughput imaging, automatically identifying regions in the tissue, and quantifying histological dye concentrations in segmented region to accurately evaluate disease marker expression.