Novel semi-automated algorithm for high-throughput quantification of adipocyte size in breast adipose tissue, with applications for breast cancer microenvironment

Figures

• 

  Figure 1.. Representation of intra-sample heterogeneity. The area enclosed in the solid-line border represents an optimal sample containing adipocytes characteristically shaped with structured cell membranes, well-defined borders and absence of cellular or foreign artefacts. Sub-optimal samples are represented in the subsequent enclosed sections. The area enclosed in the dotted line border and the area enclosed in the dashed line border depict the presence of morphological signatures and cellular artefacts. The area enclosed in the long-dash border contains poorly defined adipocyte cells with deteriorated cell membranes indicative of poor fixation or non-uniform slices that are too thin for adipose tissue

• 

  Figure 2.. Compares the calculated areas for each measurement method. (a): Measurements with the TM method are outlined in blue. (b): Measurements with HT method are outlined in red. (c): Measurements with the EA method are outlined in yellow with the calculated major axis is shown as a green line and semi-minor axis is shown in the red line. (d): AdipoCyze Graphical User Interface (GUI) shown in TM

• 

  Figure 3.. Comparison of adipocyte area calculation by method by cell. Results of analysing adipocyte size (μm2) by cell, in the order of cells measured are shown. Each data point in the bar chart represents the average of three repeated runs of for a measurement method. Measurements using the trace mode, hand trace, and ellipse approximation methods are shown in black, white, and grey bars, respectively. Error bars represent standard deviation of measurements for individual cells

• 

  Figure 4.. Adipocyte size distribution by method. Results from various methods are shown in 100% overlap to show the shape of distribution across 6 bins. Adipocyte area size distribution from the TM, HT, and EA is shown in blue check pattern, red, and yellow, respectively. Overlap between distributions is shown by secondary shades

• 

  Figure 5.. Concordance between adipocyte quantification methods was analysed and be observed in the scatter plots (a, b, c). Variables were plotted along the X, Y axes. Linear trendlines were plotted through each series. Concordance was evaluated based on slope and correlation coefficient of trendline. (a): Concordance of HT with the TM is observed from the series with white circles and black outline. (b): Concordance of HT with EA is observed in the series of black circles. (c): Concordance of the TM and the EA is observed in the series of grey circles

• 

  Figure 6.. Adipocyte analysis rate by method. The average time to measure each cell was determined amongst the different methods. Average time to measure each cell using TM, HT, and EA, are shown plotted against each cell with symbols, black triangle, white diamond, and grey circle, respectively. Linear trendlines are displayed on each series of data

• 

  Figure 7.. Multiple methods of adipocyte quantification. Challenges of intra-sample variability can be solved for utilizing multiple methods of adipocyte quantification. Due to intra-sample variability, adipocytes quantified using TM are shown outlined in blue, adipocytes quantified using EA are shown outlined in yellow, and adipocytes quantified using HT are shown outlined in red