Using discrete wavelet transformation, we devised an improved method for focus detection in an automatic measurement system for counting numerical chromosomal aberrations per nucleus. In the previous system we had developed, the focus was determined by detecting the maximum intensity. However, the focus sometimes needed to be adjusted if it happened to be on the fluorescence noise that accompanied the preparation of specimens when the noise intensity was high. By using an appropriate level of wavelet transformation of the microscopic images, the focus position for cell nuclei could be obtained even in the presence of fluorescence. In addition, in the previous system, even if undamaged nuclei were selected correctly, the system picked up only a small proportion of nuclei that satisfied the criterion, which resulted in most nuclei being excluded from measurement. This also entailed the possibility of taking up an eccentric population of nuclei, and thus reducing the reliability of the measurement results. Therefore, we have attempted to find a criterion that will allow us to select a large proportion of undamaged nuclei. Since the contour of a good nucleus is close to a circle, circles having the minimum average difference from the contour of the nucleus, the smallest standard deviation, or the largest difference were chosen. Using these parameter values, the nuclei of lymphocytes suitable for measurement were determined, and 80% of the total nuclei among those considered suitable were correctly selected. By this method, the precision and reliability of the automatic counting system were improved.
View full abstract