Micro-CT and machine learning: a high-throughput alternative to histology for follicle reserve assessment in cryopreserved ovarian tissue.

BACKGROUND

Ovarian tissue cryopreservation followed by transplantation after cancer remission is a fertility preservation strategy available for certain patient groups, such as pre-pubertal and adolescent girls, as well as adult females requiring urgent gonadotoxic therapy. Quantitative assessment of follicular density in cryopreserved cortical tissue is critical for evaluating tissue quality and estimating its reproductive potential. Conventional analysis, based on manual follicle counts in serial histological sections, is time-consuming, labor-intensive, and prone to variability from uneven follicle distribution and inconsistent tissue orientation. To address these limitations, we developed a high-throughput, automated method combining micro-CT, machine learning, and morphological analysis to quantify oocyte density and other morphological features throughout entire ovarian cortical tissue samples.

RESULTS

Three-dimensional segmentation analysis enabled quantification of oocyte density in the samples within the cortical region 1 mm below the surface epithelium. Oocytes in pediatric samples were located significantly closer to the surface compared to those in adult tissue, with median distances of 139.4 μm and 370.2 μm, respectively (P < 0.0001) and exhibited markedly higher local oocyte neighbor counts, with median values of 6 and 2 in pediatric and adult tissues, respectively (P < 0.0001), consistent with higher oocyte density and clustered spatial organization in younger individuals. Simulated histology using every 10th virtual sections -corresponding to 40 μm separated histology slices- closely approximated full-volume micro-CT estimates of oocyte density. Analysis based on only five virtual sections aligned with micro-CT data exclusively in pediatric samples with high oocyte density, whereas in adult samples it led to substantial inaccuracies in oocyte density estimation.

CONCLUSION

Micro-CT scanning combined with machine learning analysis represents a novel high-throughput and automated approach for estimating oocyte count in cryopreserved ovarian cortical samples. In addition, three-dimensional analysis offers valuable insights into oocyte localization and spatial distribution within the ovarian cortex, presenting a promising alternative to conventional histology for future clinical and research applications.