Background. While growth hormone (GH) is commonly used as an adjuvant treatment to controlled ovarian stimulation (COS) for in vitro fertilization (IVF) cycles, the data regarding its efficacy is inconsistent. Design. A retrospective matched cohort study of poor responder patients who underwent COS without the use of GH (COS-GH) subsequently followed by COS cycles that included adjuvant GH (COS+GH) treatment. Materials and Methods. A list of all patients having filled a prescription for GH from January 2018 – March 2020 was obtained. GH was administered daily at 3mg (9IU) starting on the first day of stimulation and ending on the day of trigger. Only women who had documentation of a previous cycle without the use of GH were included in the study. Results. 182 cycles (91 patients) were included in the study, and COS-GH cycles were compared to COS+GH cycles. The total dose of gonadotropins used (5757 vs 4252 mIU, p=0.002), duration of stimulation (10.4 vs 10.1 days, p=-.045), maximum Estradiol (E2) (2411 vs 1932 pg/ml, p=0.010), endometrial thickness (11.2 vs 10.6 mm, p=0.010), number of oocytes retrieved (14.2 vs 11.8, p=0.001), number of mature oocytes (11.1 vs 9.7, p=0.028), number of blastocysts (3.98 vs 2.56, p=<0.001) and number of usable blastocysts (2.5 vs 1.6, p=<0.001) were all significantly greater in the GH group. Conclusions. Adding GH to the COS protocol in poor responder patients may lead to improvements in the number of oocytes retrieved, number of mature oocytes, endometrial thickness, number of blastocysts, the number of usable blastocysts.
optical signature of each type of microscope, and the presence of foreign objects, such as a holding micropipette present in the image. If a certain optical signature or foreign artifact appeared more in the positive training class compared to the negative training class, the CNN models were found to learn these biases and give a higher score to those images regardless of the embryo morphology. With these insights, a new dataset was prepared that balanced the ratios of positive-to-negative samples for each type of microscope and for each group containing foreign objects. This provided a non-inflated total AUC of 0.72 and significantly raised the lowest per-site AUC from 0.51 to 0.61.CONCLUSIONS: There has been significant recent interest in using deep learning for analyzing images of embryos at the blastocyst stage. The blackbox nature of deep learning models such as CNNs makes it difficult to recognize when potential sources of bias have been introduced during the training process. We performed a series of experiments that identified and reduced two sources of bias, and improved per-clinic performance of the CNN. Future work will continue to search for other sources of bias and address them accordingly.IMPACT STATEMENT: Naive approaches to preparing training data for deep learning models for embryo ranking can create bias in the models. Our work illustrates the need for careful preparation of training data and monitoring of different metrics to identify and reduce potential sources of bias.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.