Figure 1.
Comparison multi-layer scanning with single-layer scanning.
Two cross sections of a LBC slide are shown. The upper one shows the multilayer scanning principle. The lines represent the particular layers. Green line parts represent in-focus regions and red line parts represent the out-of-focus regions. In multilayer scanning, the most parts of the layers are out-of-focus and thereby an unnecessary amount of data is generated. The lower cross section shows the principle of a single-layer scan. A “master-focus layer” (green line) represents the full 3D focus map of the LBC slide. In the optimal case, one focus layer would be sufficient and multi-layering would not be needed anymore or only as a supplement to cover thick cell clusters (transparent green lines).
Figure 2.
Highly detailed single-slide analysis.
(a) A schematic showing the origin of the optical z-axis; Red arrow: showing the measured distance from the objective to the measured objects. (b) A 3D graph of the focus points of two different layers which can be found on the slides. The red dots represent points focused on dust which are located on the coverslip. The blue dots are the focus points of the cell layer. The graph looks inverse comparing to the real physical location of the focus points as its origin lies in the lower left corner; (c) a 3D mesh plot of the obtained focus point data only by the cell layer of the slide shows a high degree of heterogeneity within the slides; (d) another example similar to (c) in which smaller variations in the z values were observed. The examples in (c) and (d) demonstrate that it is not possible to scan the slides as a planar mono-layer and that there is a high height variation within the slides.
Table 1.
Descriptive statistics of the focus point dataset of the particular slides showing the high variations between the z-values within and between the slides.
Figure 3.
Six different focus point images and comparison of in-focus and out-of-focus images
; (a) a valid image containing in-focus cells. Cells are stained with hematoxilin resulting in blue color; (b) a valid example showing some parts of normal stained cells and cells highlighted with a biomarker conjugated to a brown stain (DAB); (c) an invalid image with only very small objects present; (d) an invalid image with no objects present; (e) an invalid image containing in-focus artifacts; and (f) an example of an invalid out-of-focus image. (g) an example of an in-focus image, and an image which is out-of-focus (h). It is very problematic to obtain relevant image information from (h).
Table 2.
Contingency table and overall performance of the focus point analysis of 2295 focus points from 51 LBC slides.
Figure 4.
The detailed steps for whole-slide sharpness quantification;
At first, the slide is divided into 16 sub-regions. Then, cells are detected by their color values. In total 200 cells are used to quantify the sharpness of each region. For every cell, five sharpness features are computed and a support vector machine (SVM) is used to classify each cell into the in-focus (class 1) or out-of-focus(class 0) category. The percentage of in-focus cells (0–100%) is used to calculate a score for each region, and a combination of these scores is used to represent slide sharpness.
Table 3.
Confusion matrix and overall performance of the classifier used to determine the sharpness of the cell image.
Figure 5.
A simplified schematic of the complete workflow for scanning one slide.
The slide is loaded and the area to be scanned is detected automatically. Focus points are set and after autofocussing, the focus point images are analyzed. If the number of valid focus point is higher than five, the slide is scanned and its sharpness is analyzed. From the results of sharpness analysis, a decision is made whether to re-scan the slide or not. The slide is re-scanned until the quality is sufficient for further analysis.
Table 4.
Scanning duration and number of focus points for the 400 scanned LBC slides.
Table 5.
The number of slides finished after each scanning iteration of a total of 400.