The aim of this study was to utilize bitmap segmentation of ultrasound images to enhance the consistency and accuracy of determining the chemical composition of ram testicular parenchyma based on its first-order echo- textural characteristics, including mean numerical pixel intensity (MPI), standard deviation of numerical pixel values (mean pixel heterogeneity-MPH), and pixel frequency distribution (mean pixel concentration-MPC). Testes (n = 10) were obtained post-mortem from ten different, sexually mature Karakul rams and scanned using an 8-MHz linear-array transducer in both longitudinal (Long) and transverse (Trans) planes. The crude fat, protein, and moisture contents of testicular tissue samples were measured using validated laboratory techniques. Digital ultrasonograms of the testicular parenchyma were normalized and converted to bitmap files using commercially available ImageProPlus® analytical software. Two segmentation approaches were compared to assess their ability to detect quantitative correlations between echotextural characteristics and proximate chemical composition of testicular tissue: echointensity (EI) banding and algorithmic segmentation of the bit- maps. Using 50- or 25-pixel intensity bands, only two Long EI bands and five Trans EI bands showed significant correlations between MPI, MPH, or MPC and testicular tissue constituents. The accuracy of predicting testicular composition using algorithmically identified pixel intensity clusters ranged from 85.41 ± 3.67 % (MPI-fat) to 99.69 ± 0.12 % (MPC-moisture) for Long images, and from 78.31 ± 5.24 % (MPI-fat) to 99.76 ± 0.06 % (MPH- moisture) for Trans images. These results suggest that computerized detection of specific pixel intensity ranges offers an optimal method for determining the correlations between first-order echotextural characteristics and the chemical composition of ram testes.

Keywords