Predicting pregnancy test results after embryo transfer by image feature extraction and analysis using machine learning

Abstract

Assessing the viability of a blastosyst is still empirical and non-reproducible nowadays. We developed an algorithm based on artificial vision and machine learning (and other classifiers) that predicts pregnancy using the beta human chorionic gonadotropin (b-hCG) test from both the morphology of an embryo and the age of the patients. We employed two high-quality databases with known pregnancy outcomes (n = 221). We created a system consisting of different classifiers that is feed with novel morphometric features extracted from the digital micrographs, along with other non-morphometric data to predict pregnancy. It was evaluated using five different classifiers: probabilistic bayesian, Support Vector Machines (SVM), deep neural network, decision tree, and Random Forest (RF), using a k-fold cross validation to assess the model's generalization capabilities. In the database A, the SVM classifier achieved an F1 score of 0.74, and AUC of 0.77. In the database B the RF classifier obtained a F1 score of 0.71, and AUC of 0.75. Our results suggest that the system is able to predict a positive pregnancy test from a single digital image, offering a novel approach with the advantages of using a small database, being highly adaptable to different laboratory settings, and easy integration into clinical practice.

Embryo selection with artificial intelligence: how to evaluate and compare methods?

“…1 Only aneuploid miscarriages (confirmed with genetic testing of chorionic villus samples) were included as negative live births 2 Negative fetal heartbeat was assumed for all non-transferred embryos that had "failed or abnormal fertilization, grossly abnormal morphology or aneuploidy from preimplantation genetic testing" on low quality embryos at different developmental stages, in order to ensure that the evaluation data is representative of prospective use [2]. Other models seek to differentiate between previously transferred embryos [12,14,[20][21][22]24]. When only evaluating on transferred embryos, such models assume that an embryologist first preselects potentially transferable embryos (e.g., day 5 blastocysts).…”

“…AUC has been falsely accused of being influenced by class imbalance [12,21] with the conclusion that "the metric cannot be trusted in highly unbalanced data" [8]. However, as described above, AUC is independent of prevalence and thus not influenced by unbalanced datasets.…”

A Deep Learning Approach for Prediction of IVF Implantation Outcome from Day 3 and Day 5 Time-Lapse Human Embryo Image Sequences

“…The second category attempts to correlate between embryos physical attributes and their outcome after the embryos are transferred. The recorded outcome can be either in the form of implantation [17], or livebirth [18,19,20]. Such an actual outcome is used as the ground truth (GT), and AI models are created to predict the correct outcome through analyzing embryo images.…”

“…These incubators have enabled continuous monitoring of embryos' development. Unfortunately, most AI-based developed systems only evaluate embryos quality based on a single shot of its blastocyst stage [18,19,11,12]. Only a few methods have utilized time-lapse image sequences [13,20].…”

“…When an algorithm focuses only on an embryo at one specific development stage, only the visual attributes existing at that stage will be processed, and perhaps the duration of the process will be dismissed. In [18,19,11,12], embryos are evaluated only based on a single frame at their blastocyst stage. Other vital information, such as the length of time to reach the blastocyst stage, will be dismissed entirely.…”