Performance of a deep learning based neural network in the selection of human blastocysts for implantation
- Massachusetts General Hospital, Harvard Medical School, United States
- Brigham and Women's Hospital, Harvard Medical School, United States
- Extend Fertility, United States
- San Diego Fertility Center, United States
- Colorado Center for Reproductive Medicine IVF Laboratory Network, United States
- University of Wisconsin, United States
Abstract
Deep learning in in-vitro fertilization is currently being evaluated in the development of assistive tools for the determination of transfer order and implantation potential using time-lapse data collected through expensive imaging hardware. Assistive tools and algorithms that can work with static images, however, can help in improving the access to care by enabling their use with images acquired from traditional microscopes that are available to virtually all fertility centers. Here, we evaluated the use of a deep convolutional neural network (CNN), trained using single timepoint images of embryos collected at 113 hours post-insemination, in embryo selection amongst 97 clinical patient cohorts (742 embryos) and observed an accuracy of 90% in choosing the highest quality embryo available. Furthermore, a CNN trained to assess an embryo’s implantation potential directly using a set of 97 euploid embryos capable of implantation outperformed 15 trained embryologists (75.26% vs. 67.35%, P<0.0001) from 5 different fertility centers.
Data availability
Patients did not explicitly consent to their data being made public and access is therefore restricted. Requests for the anonymized data should be made to Charles Bormann (cbormann@partners.org) and Hadi Shafiee (hshafiee@bwh.harvard.edu). Requests will be reviewed by a data access committee, taking into account the research proposal and intended use of the data. Requestors are required to sign a data-sharing agreement to ensure patients' confidentiality is maintained prior to the release of any data.
Article and author information
Author details
Funding
National Institutes of Health (R01AI138800)
- Hadi Shafiee
Brigham and Women's Hospital (Precision Medicine Program)
- Hadi Shafiee
Partners Healthcare (Innovation Discovery Grant)
- Charles L Bormann
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Ethics
Human subjects: Embryo image/video data collected from patients were used in this study with an institutional review board approval (IRB#2017P001339) We used 3,469 recorded videos of embryos collected from patients with informed consent for research and publication, under an institutional review board approval for secondary research use.
Copyright
© 2020, Bormann et al.
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 5,073
- views
-
- 855
- downloads
-
- 92
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Performance of a deep learning based neural network in the selection of human blastocysts for implantation
eLife 9:e55301.
https://doi.org/10.7554/eLife.55301
Further reading
-
- Immunology and Inflammation
- Medicine
Preterm infants are susceptible to neonatal sepsis, a syndrome of pro-inflammatory activity, organ damage, and altered metabolism following infection. Given the unique metabolic challenges and poor glucose regulatory capacity of preterm infants, their glucose intake during infection may have a high impact on the degree of metabolism dysregulation and organ damage. Using a preterm pig model of neonatal sepsis, we previously showed that a drastic restriction in glucose supply during infection protects against sepsis via suppression of glycolysis-induced inflammation, but results in severe hypoglycemia. Now we explored clinically relevant options for reducing glucose intake to decrease sepsis risk, without causing hypoglycemia and further explore the involvement of the liver in these protective effects. We found that a reduced glucose regime during infection increased survival via reduced pro-inflammatory response, while maintaining normoglycemia. Mechanistically, this intervention enhanced hepatic oxidative phosphorylation and possibly gluconeogenesis, and dampened both circulating and hepatic inflammation. However, switching from a high to a reduced glucose supply after the debut of clinical symptoms did not prevent sepsis, suggesting metabolic conditions at the start of infection are key in driving the outcome. Finally, an early therapy with purified human inter-alpha inhibitor protein, a liver-derived anti-inflammatory protein, partially reversed the effects of low parenteral glucose provision, likely by inhibiting neutrophil functions that mediate pathogen clearance. Our findings suggest a clinically relevant regime of reduced glucose supply for infected preterm infants could prevent or delay the development of sepsis in vulnerable neonates.
-
- Medicine
- Microbiology and Infectious Disease
Persisting HIV reservoir viruses in resting CD4 T cells and other cellular subsets are a barrier to cure efforts. Early antiretroviral therapy (ART) enables post-treatment viral control in some cases, but mechanisms remain unclear. We hypothesised that ART initiated before peak viremia impacts HIV-1 subtype C reservoirs. We studied 35 women at high risk of infection from Durban, South Africa, identified with hyperacute HIV by twice-weekly HIV-RNA testing. Participants included 11 starting ART at a median of 456 (297–1203) days post-onset of viremia (DPOV) and 24 at 1 (1–3) DPOV. Peripheral blood mononuclear cells (PBMCs) were used to measured total HIV-1 DNA by droplet digital PCR (ddPCR) and sequence viral reservoir genomes by full-length proviral sequencing (FLIP-seq). ART during hyperacute infection blunted peak viremia (p<0.0001), but contemporaneous total HIV-1 DNA did not differ (p=0.104). Over 1 year, a decline of total HIV-1 DNA was observed in early treated persons (p=0.0004), but not late treated. Among 697 viral genome sequences, the proviral genetic landscape differed between untreated, late treated, and early treated groups. Intact genomes after 1 year were higher in untreated (31%) versus late treated (14%) and early treated (0%). Treatment in both late and early infection caused more rapid decay of intact (13% and 51% per month) versus defective (2% and 35%) viral genomes. However, intact genomes persisted 1 year post chronic treatment but were undetectable with early ART. Early ART also reduced phylogenetic diversity of intact genomes and limited cytotoxic T lymphocyte immune escape variants in the reservoir. Overall, ART initiated in hyperacute HIV-1 subtype C infection did not impact reservoir seeding but was associated with rapid intact viral genome decay, reduced genetic complexity, and limited immune escape, which may accelerate reservoir clearance in combination with other interventional strategies.
