415 USE OF COMPUTER-ASSISTED ULTRASOUND IMAGE ANALYSIS IN EMBRYO RECIPIENT SELECTION

Abstract

Ultrasonography has been extensively used in the study of reproductive function in areas such as antral follicular dynamics, ovulation, luteal function, and early pregnancy. However, most studies with ultrasound evaluation are limited to measurement of diameter, length, or area of internal structures, and few groups use differences in image density (echotexture) as an evaluation parameter. For corpora lutea, echo-texture differences reflect variations in vascularization and luteal cell proportions. The aim of the present study was to evaluate image analysis as a tool for embryo recipient selection. Grades I or II embryos produced by conventional superovulation (n = 55) or IVF (n = 87) were nonsurgically transferred to primiparous cows or heifer recipients (n = 142). Estrous synchronization was performed using the heat-synch protocol (Day 0: CIDR^®; Day 5: 400 IU of eCG; Day 7: 500 µg of cloprostenol plus CIDR removal; Day 9: 2 mg of estradiol benzoate). Sonographic corpora lutea images were obtained on the day of embryo transfer (Day 7), using a portable ultrasound device (Aloka SSD 500; Aloka Co., Sao Paulo, Brazil) equipped with a linear rectal 5 MHz probe. Images were digitalized in .TIFF format at a resolution of 1500 × 1125, using a video capture board (Pinnacle DC10; Pinnacle Systems, Mountain View, CA, USA). A representative elementary area (REA) of 2.704 pixels (25 mm²) was defined in the luteal tissue, using the criteria proposed by Van den Bygaart et al. (1999 Can. J. Soil Sci. 79, 149–160). Image analysis was performed using a custom software package (QuantPro^®). Each image dot (pixel) received a numeric value ranging from 0 (black) to 255 (white). Data from cows and heifers later diagnosed as pregnant or not in the groups receiving SOV or IVF embryos were analyzed by ANOVA, and results are presented as means ± SEM. As expected, the pregnancy rate was higher for conventional than for IVF embryos (58.2% vs. 31.0%; P < 0.05). There was a low correlation (R = 0.33) between luteal tissue area and pixel value for Day 7 corpora lutea. There was no difference (P > 0.05) in luteal tissue area or echotexture between pregnant and non-pregnant animals in the groups receiving superovulation embryos (3.12 ± 0.26 cm² vs. 2.63 ± 0.25 cm², and 78.03 ± 2.25 vs. 79.73 ± 3.73, respectively) and IVF embryos (3.25 ± 0.24 cm² vs. 3.03 ± 0.14 cm², and 74.81 ± 1.93 vs. 70.82 ± 1.62, respectively). However, in the IVF embryo group, no pregnancy was established in cows bearing a CL with mean pixel value lower than 60.25 or greater than 89.27 (total values ranged from 44.72 to 99.79). Corpora lutea image analysis, using the REA proposed in this approach, had limited value to predict pregnancy rate of embryo recipients. Further studies will investigate the accuracy of the REA established and other image characteristics, including the pixel distribution pattern.