Table 1.
Input parameters of the algorithm.
Table 2.
Output features of the algorithm.
Fig 1.
MIP and corresponding height view.
The Maximum Intensity Projection (MIP) and the height view of an image are shown (a-b) and for a single xz-plane from the image (c) it is indicated (by red/white squares) where the pixels originated from, that is, which z-coordinates resulted in the highest intensity. The corresponding z-values get mapped to the height view, while the pixel intensities are mapped to the MIP. When two spheroids are overlapping in the lateral direction, the MIP will show the spheroid with the highest intensity.
Fig 2.
Segmentation of the spheroids in 2D.
(a): Part of the maximum intensity projection image of the RFP channel of a 3D image stack, (b): corresponding height view, (c): After applying a range filter on the height view, giving the local variance in the z-depth, where bright (dark) pixels indicate background (spheroid) regions, (d): histogram of (a), showing that the foreground cannot distinguished clearly, (e): histogram of (c), showing that separation between foreground and background is possible. (f) after segmentation.
Fig 3.
(a) The segmented spheroids in 2D, overlaid on the MIP. (b-c) Ellipsoids fitted to the spheroid mask in (a). The center slice projections of the spheroids from (b) the top (the xy-plane) and (c) the side (the xz-plane), are shown overlaid with the fitted ellipsoids.
Fig 4.
Light attenuation in multi-cellular spheroids.
(a) Top view from the xy-plane of a spheroid with an ellipsoid fitted, where the RFP (561 nm), Hoechst (405 nm), and EdU (640 nm) channels are shown. (b) The same spheroid, in a side view (xz-plane). No signal is detected from the lower part of the spheroid (assuming that the spheroid is of an ellipsoidal shape). (c) Parameters of the spheroid derived from the vertical profile curve of the RFP signal through the ellipsoid center: top z-coordinate, maximum intensity, analyzable depth (corresponding to the user-defined minimum intensity percentage).
Fig 5.
Validation of the ellipsoid segmentation algorithm.
(a-c) The classification of spheroids being obscured by others in the GT data is compared with the automated identification of those spheroids based on the circularity of the 2D spheroid masks. (a) illustrates a typical situation where the shape of the 2D spheroid mask is non-elliptical because the spheroid is overlapping with another (brighter) one, the green dashed line shows the GT contour of the spheroid, while the solid contours result from automatic segmentation. (b) shows the spheroid circularity resulting from the automatic segmentation as function of the spheroid class in the GT, where the different classes are given by the cartoons on the x-axis representing well separated (red), adjacent (blue), or overlapping (dark green) spheroids. (c) shows the percentage of correctly segmented pixels by the automatic segmentation as function of the circularity for the class of well separated (red) and overlapping (green) spheroids. (d) shows spheroids of different sizes that are categorized according to their analyzable region: the entirety, more than half, or less than half of the spheroid is analyzable (visibility category). Above the bars the number of spheroids per visibility category is shown. The labels large (blue), medium (red), and small (green) correspond to spheroids with a number of cells larger than 200, between 50 and 200, and smaller than 50, respectively.
Fig 6.
Proliferation analysis example data.
(a): Scheme of the labeling of the cells and dyes used. (b-c): Image of a 3D in vitro cancer cell culture labeled / stained with EdU, RFP, GFP and Hoechst. The maximum intensity projection is shown.
Fig 7.
Comparison of the proposed analysis method with a baseline 2D MIP analysis method.
Spheroid cell culture proliferation, quantified by EdU positive cells, from a 3D homogeneous spheroid culture treated with a cytotoxic compound (Docetaxel, 1e-8 M) and a cytostatic compound (MDV-3100, 1e-7 M). In (a) the spheroid volume is shown, and in (b) the number of EdU positive cells per spheroid. In (c) the total number of EdU positive cells per image volume is shown, where the size of the dots represent the total foreground / background ratio of the MIP of the image stack. In (d), the EdU positive cell count, based both on the proposed and a 2D MIP analysis method, is compared. The normalized difference in EdU positive cells, weighted with the spheroid volume, is shown.