High-contrast imaging of moving targets

Summary form only given. The imaging of cells flowing through blood vessels can be a useful diagnostic tool. However, live samples are usually transparent and contrast is poor with bright-field microscopy. A variety of methods are available to improve contrast including; phase contrast, differential interference contrast, fluorescent labelling and surface plasmon resonance imaging. These methods are not all amenable to automated cell detection and analysis, however, because the cell is not easily separable from the background. Here we compare the contrast improvements achieved using three differential imaging methods. These methods remove the invariant background, including the vessels and surrounding tissue, and identify the changes in successive frames in a sequence. They can be applied either online or offline and no preparation of the microscope or samples is necessary.Blood vessels in a live water flea (Daphnia) were imaged at 25 fps using a CMOS camera. The resulting image sequence was convolved with different finite impulse response (FIR) filters; a simple successive-subtraction, a high-pass and a differentiator. Contrast was then defined as the ratio of the target coefficient-of-variance (CV) to the background CV. The simple filter (h=[-l 1]) gave the best performance with a CV ratio of 7.6; while the differentiator achieved 4.9 and the high-pass filter 1.8, despite being of a higher order. This quick comparison demonstrates that a simple subtraction filter can give superior contrast improvement with the advantage of reduced computational demand over more complex FIR filters. It also readily lends itself to applications that require automated cell detection and analysis.