Fig 1.
Challenges associated with plasma cell segmentation; numbers in boxes indicate the following image regions: nucleus of plasma cells,
cytoplasm of plasma cells,
unstained cells, and
background.
Three challenges are highlighted via this Fig: 1) At times, the color difference between the cytoplasm with the adjacent background is less; 2) Plasma cells may be clustered together and hence, segmentation of the overlapping/touching cells is required; and 3) there may be more than one type of stained and unstained cells posing difficulty in extracting plasma cells of interest.
Fig 2.
Histogram of four image regions (nucleus, cytoplasm, background, and unstained cells) in:(a)-(c) RGB, (d)-(f) HSV, and (g)-(i): Lab color spaces; Corresponding fitted probability density functions (PDFs): (a′)-(c′) RGB, (d′)-(f′) HSV, and (g′)-(i′): Lab color spaces.
Fig 3.
Contrast stretching followed by unstained cell removal, image patch size 2560 × 1920 this figure shows (a) an original image that is (b) contrast stretched such that 1% of lower and higher intensity values are saturated to 0 and 255, respectively.
From the probability density functions of the four regions of interest (nucleus of PC, cytoplasm of PC, unstained cells, and background) of the resulting contrast stretched images, it is observed that nucleus and cytoplasm of plasma cells have maximum separability with unstained cells in H color channel. Unstained cells are removed by replacing intensity of pixels having values less than 120 in the H-channel with the background pixel intensity leading to (c).
Table 1.
Bhattacharyya distance calculated between different image regions using the ground truth data.
Table 2.
Weights of each ROI for level set equations.
Fig 4.
Schematic diagram of Steps 1 to 3 of the proposed method of PCSeg tool: Regions of interest (ROI) are: Nuclei of Plasma Cells (Ω11), Cytoplasm of Plasma Cells (Ω10), unstained cells (Ω01), and background (Ω00).
Fig 5.
Output of the modified multiphase level set method after Step-3 on an image patch.
It was noted that some unwanted stained cells (e.g. lymphocytes) were segmented to final output. In addition, some small disconnected components that were noisy patches owing to faulty manual staining were also captured by the multiphase level set. Also plasma cell clusters were not segmented.
Fig 6.
Output of cluster cell segmentation on an image patch Subfigure (a) shows the level set output wherein red boundary shows nucleus of PC being captured by levelset and green boundary shows cytoplasm of PC being captured. However, the cluster of cells are not segmented. (b) shows nuclei identified by levelset phase Ω11 and the problem therein of some extra mask of cytoplasm in nucleus. (d) shows cluster cell segmentation using nuclei mask identified by level set Ω11 in (b). One cell is falsely rejected. (c) shows nuclei identified by k-means on the levelset phase Ω11 and (e) shows correct cluster cell segmentation using the mask of (c).
Fig 7.
Schematic diagram of cluster cell segmentation (Step 4 of PCSeg tool) using watershed and circle Hough transform (CHT).
Fig 8.
Running times on CPU and GPU.
Fig 9.
Qualitative comparison of MM cell segmentation using different methods: (a) Gold standard (showing cells of interest with white outlines), (b) k-means, (c) Chan-Vese active contour method [16], (d) Chan-Vese multiphase method [29], (e) Saeedizadeh et al. [24] method, (f) PCSeg Tool-1, and (g) PCSeg Tool-2. All white outlines in (b)-(g) denote the outlines of regions segmented out. These regions are required to be compared with the regions contained in the Gold Standard shown in (a).
Table 3.
Quantitative results on plasma cell segmentation.
Table 4.
Statistical results on a total of 364 (260+102) plasma cells.
Fig 10.
Qualitative comparison of MM cell segmentation using different methods over five images: (a) Gold standard (showing cells of interest with white outlines), (b) Saeedizadeh et al. [24] method, (c) PCSeg Tool-1, and (d) PCSeg Tool-2. White outlines in all figures (b)-(d) denote the outlines of regions segmented out. These regions are required to be compared with the regions contained within the white boundaries in Gold Standard shown in (a).