Abstract
Study question
Can the algorithm used by EmbryoScopePlus software predict implantation and clinical pregnancy in women of different age groups on fresh transfer?
Summary answer
The embryo score generated by KIDScoreD5 is highly related to the rates of implantation and clinical pregnancy in fresh transfers in women of different age.
What is known already
Artificial Intelligence algorithms use statistics to find patterns in large amounts of data and describe a non-biased approach to multiparameter analysis. Several algorithms have been described, but none has been adopted for universal use. KIDScoreD5 is the algorithm included in the EmbryoScopePlus system and classifies embryos according to the cleavage times and morphology of the blastocyst. Version 3, more current, includes the annotations of the number of pronuclei, the time of division for 2, 3, 4 and 5 cells, time to start of blastulation, and morphology of the Internal Cell Mass and trophectoderm.
Study design, size, duration
Retrospective study evaluated 86 embryos from January to December 2019 at the Reproferty clinic, grown at EmbryoScopePlus and transferred fresh on the fifth day of embryo development. The morphological and morphokinetic parameters were automatically evaluated by the software and in case of any mistake, they were manually corrected. The embryos were evaluated by KIDScoreD5 v3 in different scores from 0.0 to 9.9 and divided into 4 groups (0.0–2.5; 2.6–5.0; 5.1–7.5; 7.6 –9.9).
Participants/materials, setting, methods
The inclusion criterion was transfer of a single embryo with 1 gestational sac and positive FHB and transfer of two embryos with 2 gestational sac and positive FHB. Patients with progesterone on the trigger day ≥ 1.5ng/mL and/or with endometrium ≤7mm were excluded. The implantation and clinical pregnancy rates were calculated according to age group, G1: ≤35 years; G2: between 36 and 39 years old; G3: ≥40 years, within the embryo classification.
Main results and the role of chance
For patients in group 1 (n = 31 embryos), 33.4% of the embryos were classified between 2.6–5.0; 69.20% of embryos with scores between 5.1–7.5 and 57.10% of embryos with scores between 7.6–9.9, with 100% of embryos that implanted, regardless of classification, resulting in clinical pregnancy . For group 2 (n = 35 embryos), they only showed an implantation rate for embryos where the scores were 5.1–7.5 (33.4%) and 7.6 - 9.9 (71.4%) , with 100% being the clinical pregnancy rate in these groups. For patients in group 3 (n = 24 embryos), we also observed implantation only in groups of embryos with a score of 5.1–7.5 (37.5%) and 7.6–9.9 (18.5%) , but the clinical pregnancy rate was lower when compared to the other age groups of the patients, with 33.5% for embryos having a score between 5.1–7.5 and 50% for the group 7.6–9.9. Regarding the average score given by the classification of KIDScore Day 5 v. 3 for embryos that implanted, for patients aged 35 years or less, the average was 6.92; for patients between 36 and 39 years old, the average was 8.06 and for patients aged 40 years or older, the average was 7.32.
Limitations, reasons for caution
This project is limited because it is a retrospective study and evaluated embryos from a single breeding center. Multicenter and prospective studies are necessary to validate the universal use of the KIDScoreD5 v3 algorithm in time-lapse incubators.
Wider implications of the findings: The study showed the ability of KIDScoreD5 v3 to assist the embryologist in deciding which embryo to transfer fresh, according to the patient’s age, in addition to the software being effective in automatic annotation of morphological and morphokinetic parameters. Validating an algorithm universally will improve embryonic selection.
Trial registration number
Not applicable
Interobserver Variability in the Differential Diagnosis of Benign Bone Tumors and Tumor-like Lesions
