Supplementary AIVE descriptions and demonstrations. (A) Signal matrix depicting how the input values for a voxel contribute to the output value during AIVE (red line indicates approximate location of threshold used for 3D rendering and segmentation). (B) 3D examples of other various categories of electron signal processed via AIVE, as detected in the test dataset shown in Fig. 1. (C) Quantification of the average ΔV per slice for each two-way comparison made between the model predictions from Fig. 1 F after the application of a 10-nm Gaussian blur (3D renders of corresponding data also provided in Fig. S2 B). (D) 3D renderings indicating the position of each static 3D probe used for mitochondrial distance measurements in Fig. 3, viewed from the front and above (membranes nearest to each 3D probe are indicated by red shading). See Video 2 for rotation animation showing probe positions. (E) Maximum value projection of AIVE-processed data for the nucleus used in Fig. 3, depicting the position of each nuclear pore used for analysis in Fig. 3, N–T. (F) Averaged value projections of the unprocessed EM data after spatial alignment for analyses shown in Fig. 3, N–T. (G–I) A mathematical description of the smallest nonzero distance that can be measured between any two objects segmented by conventional binarization methods, after 3D reconstruction via the marching cubes algorithm. (G) Unique cases of cubes triangulated via the marching cubes algorithm with multiple non-connected triangles (Lorensen and Cline, 1987), which is the minimum requirement for representing two separate objects (cases 3, 4, and 7 omitted for brevity). Circles at the cube corners represent the eight voxel values used by the marching cubes algorithm to triangulate surfaces in each case. For binarized input data, the triangulated surface vertices are generated at the exact mid-point between voxel centers, which are equidistant if the voxels are spatially isometric. (H) If the spatial scale of one voxel dimension (“c”) exceeds the other two (“a” and “b”), then the separation distance between vertices in that dimension will always be greater; by extension, the shortest nonzero distance between surfaces cannot occur in that dimension, so it can be ignored. (I) The shortest possible measurable nonzero distance between two binarized objects in an anisotropic voxel grid can therefore be determined by the Pythagorean theorem, in the 2D plane defined by dimensions a and b. Interquartile range for chart is indicated by box, median is indicated by horizontal lines, minimum and maximum are indicated by whiskers. Scale bars; B, 200 nm; E, 1,000 nm; F, 100 nm.
Figure S1.

Supplementary AIVE descriptions and demonstrations. (A) Signal matrix depicting how the input values for a voxel contribute to the output value during AIVE (red line indicates approximate location of threshold used for 3D rendering and segmentation). (B) 3D examples of other various categories of electron signal processed via AIVE, as detected in the test dataset shown in Fig. 1. (C) Quantification of the average ΔV per slice for each two-way comparison made between the model predictions from Fig. 1 F after the application of a 10-nm Gaussian blur (3D renders of corresponding data also provided in Fig. S2 B). (D) 3D renderings indicating the position of each static 3D probe used for mitochondrial distance measurements in Fig. 3, viewed from the front and above (membranes nearest to each 3D probe are indicated by red shading). See Video 2 for rotation animation showing probe positions. (E) Maximum value projection of AIVE-processed data for the nucleus used in Fig. 3, depicting the position of each nuclear pore used for analysis in Fig. 3, N–T. (F) Averaged value projections of the unprocessed EM data after spatial alignment for analyses shown in Fig. 3, N–T. (G–I) A mathematical description of the smallest nonzero distance that can be measured between any two objects segmented by conventional binarization methods, after 3D reconstruction via the marching cubes algorithm. (G) Unique cases of cubes triangulated via the marching cubes algorithm with multiple non-connected triangles (Lorensen and Cline, 1987), which is the minimum requirement for representing two separate objects (cases 3, 4, and 7 omitted for brevity). Circles at the cube corners represent the eight voxel values used by the marching cubes algorithm to triangulate surfaces in each case. For binarized input data, the triangulated surface vertices are generated at the exact mid-point between voxel centers, which are equidistant if the voxels are spatially isometric. (H) If the spatial scale of one voxel dimension (“c”) exceeds the other two (“a” and “b”), then the separation distance between vertices in that dimension will always be greater; by extension, the shortest nonzero distance between surfaces cannot occur in that dimension, so it can be ignored. (I) The shortest possible measurable nonzero distance between two binarized objects in an anisotropic voxel grid can therefore be determined by the Pythagorean theorem, in the 2D plane defined by dimensions a and b. Interquartile range for chart is indicated by box, median is indicated by horizontal lines, minimum and maximum are indicated by whiskers. Scale bars; B, 200 nm; E, 1,000 nm; F, 100 nm.

This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal