High-throughput measurement of adipocyte size with open-source software using whole-slide adipose tissue images

Abstract

The aim of this study was to create and validate a high-throughput method based on open-source software for the measurement of adipocyte diameters in white adipose tissue histological sections. Human omental and subcutaneous adipose tissue samples collected during bariatric surgery were used to prepare hematoxylin and eosin-stained histological slides. Digital images were acquired. Adipocyte diameters were measured both manually and with an automated procedure created using ImageJ. Comparative analysis of our automated method with the manual measurement and associations of the mean adipocyte diameters with cardiometabolic markers were used to validate our method. A total of 377 adipose samples (190 participants) were included in the analysis. Pearson correlation of mean adipocyte diameters shows a strong linear relationship between methods (r=0.88, p<0.0001). The average diameter measured with the automated method was significantly smaller (8.1{+/-}5.3{micro}m difference, p<0.0001) compared to the manual method, likely reflecting bias in selecting the cells measured with the manual approach. Pearson correlation analyses between mean omental adipocyte diameters and markers of cardiometabolic risk show that the diameters of both methods are significantly associated with the same parameters (fasting concentrations of TG, HDL-Chol, homeostasis model assessment insulin resistance, and visceral adiposity index values) with no significant differences between methods. There were also no significant differences between the manual and automated method regarding the correlations between mean subcutaneous adipocyte diameters and anthropometric or metabolic markers. In conclusion, we have created and validated a rapid automated method based on open-source software to measure adipocyte diameters from whole-slide adipose tissue images.