Method and system for performing non-invasive genetic testing using an artificial intelligence (AI) model
An Artificial Intelligence (AI) based computational system is used to non-invasively estimate the presence of a range of aneuploidies and mosaicism in an image of embryo prior to implantation. Aneuploidies and mosaicism with similar risks of adverse outcomes are grouped and training images are labelled with their group. Separate AI models are trained for each group using the same training dataset and the separate models are then combined, such as by using an Ensemble or Distillation approach to develop a model that can identify a wide range of aneuploidy and mosaicism risks. The AI model for a group is generated by training multiple models including binary models, hierarchical layered models and a multi-class model. In particular the hierarchical layered models are generated by assigning quality labels to images. At each layer the training set is partitioned in the best quality images and other images. The model at that layer is trained on the best quality images, and the other images are passed down to the next layer and the process repeated (so the remaining images are separated into next best quality images and other images). The final model can then be used to non-invasively identify aneuploidy and mosaicism and associated risk of adverse outcomes from an image of an embryo prior to implantation.
1 . A method for computationally generating an aneuploidy screening Artificial Intelligence (AI) model for screening embryo images for the presence of aneuploidy, comprising:
defining a plurality of chromosomal group labels, wherein each group comprises one or more distinct aneuploidies comprising distinct genetic alterations or chromosomal abnormalities;
generating a training dataset from a first set of images, wherein each image comprises an image of an embryo captured after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present in at least one cell of the embryo, and the training dataset includes images labelled with each of the chromosomal groups;
generating a test dataset from a second set of images, wherein each image comprises an image of an embryo taken after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present, and the test dataset includes images labelled with each of the chromosomal groups;
separately training at least one chromosomal group AI model for each chromosomal group using the training dataset for training all models, wherein each chromosomal group AI model is trained to identify morphological features in images labelled with the associated chromosomal group label, and/or training at least one multi-group AI model on the training data wherein each multi-group AI model is trained to independently identify morphological features in images labelled with each of the associated chromosomal group labels to generate a multi-group output on an input image to indicate whether at least one aneuploidy associated with each of the chromosomal group is present or absent in the image;
selecting a best chromosomal group AI model for each of the chromosomal groups or a best multi-group AI model using the test dataset;
and deploying the selected AI model to screen embryo images for the presence of one or more aneuploidies.
2 . The method as claimed in claim 1 , wherein the step of separately training at least one chromosomal group AI model for each chromosomal group and/or training at least one multi-group AI model comprises training a hierarchical layered model, wherein training a hierarchical layered model comprises:
training a hierarchical sequence of layered models, wherein at each layer, images associated with a chromosomal group are assigned a first label and are trained against a second set of images, wherein the second set of images are grouped based on a maximum level of quality, and at each sequential layer the second set of images are subset of images from the second set in the previous layer having a lower quality than the maximum quality of the second set in the previous layer.
3 . The method as claimed in claim 2 , wherein training a hierarchical layered model comprises:
allocating a quality label to each image in the plurality of images, wherein the set of quality labels comprises a hierarchical set of quality labels comprising at least “viable euploid embryos”, “euploid non-viable embryos”, “non-serious aneuploid embryos”, and “serious aneuploid embryos”;
training a top layer model by dividing the training set into a first quality dataset with a “viable euploid embryos” label and another dataset comprising all other images, and training a model on images labelled with the chromosomal group and images in the first quality dataset;
sequentially training one or more middle layer models, wherein at each middle layer a next quality level dataset is generated from selecting images with labels with the highest quality label in the other dataset, and a model is trained on images labelled with the chromosomal group and images in the next quality dataset; and
training a base layer model on images labelled with the chromosomal group and images in the other dataset from the previous layer.
4 . The method as claimed in claim 3 , wherein after training a first base level model for a first chromosomal group, for each other chromosomal group training a hierarchical layered model comprises training the other chromosomal group against the other dataset used to train the first base level model.
5 . The method as claimed in claim 2 wherein the step of separately training at least one chromosomal group AI model for each chromosomal group further comprises training one or more binary models for each chromosomal group comprising:
labelling images in the training dataset with a label matching the chromosomal group with a present label, and labelling all other images in the training set with an absent label, and training a binary model using the present and absent labels to generate a binary output on an input image to indicate whether the a chromosomal abnormality associated with the chromosomal group is present in the image.
6 . The method as claimed in claim 2 wherein the hierarchical layered models are each binary models.
7 . The method as claimed in claim 1 , wherein each chromosomal group further comprises a plurality of mutually exclusive aneuploidy classes wherein the probabilities of the aneuploidy classes within a chromosomal group sum to 1, and one or more of the AI models is a multi-class AI model trained to estimate the probability of each aneuploidy class within a chromosomal group.
8 . The method as claimed in claim 7 , wherein the aneuploidy classes comprise (“loss”, “gain”, “duplication”, “deletion”, “normal”).
9 . The method as claimed in claim 1 , further comprising:
generating an Ensemble model for each chromosomal group comprising:
training a plurality of final models, wherein each of the plurality of final models is based on the best chromosomal group AI model for the respective group, and each of the plurality of final models is trained on the training dataset with a different set of initial conditions and image ordering; and
combining the plurality of trained final models according to an ensemble voting strategy.
10 . The method as claimed in claim 1 , further comprising:
generating a distillation model for each chromosomal group comprising:
training a plurality of teacher models, wherein each of the plurality of teacher models is based on the best chromosomal group AI model for the respective group, and each of the plurality of teacher models is trained on at least part of the training dataset with a different set of initial conditions and image ordering; and
training a student model using the plurality of trained teacher models on the training dataset using a distillation loss function.
11 . The method as claimed in claim 1 , further comprising:
receiving a plurality of images, each image comprises an image of an embryo taken after in-vitro fertilization and one or more aneuploidy outcomes;
separating the plurality of images into the first set of images and the second set of images, and assigning one or more chromosomal group labels to each image based on the associated one or more aneuploidy outcomes, wherein the first set of images and the second set of images have similar proportions of each of the chromosomal group labels.
12 . The method as claimed in claim 1 , wherein each group comprises a plurality of distinct aneuploidies with similar risks of adverse outcomes.
13 . The method as claimed in claim 12 , wherein the plurality of chromosomal group labels comprises at least a low risk group, and a high risk group.
14 . The method as claimed in claim 13 wherein the low risk group comprises at least chromosomes 1, 3, 4, 5, 17, 19, 20 and ‘47,XYY’ and the high risk group comprises a least chromosomes 13, 16, 21 and ‘45,X’, ‘47,XXY’ and ‘47, XXX’.
15 . The method as claimed in claim 1 wherein the images are captured within 3 to 5 days post fertilization.
16 . The method as claimed in claim 1 , wherein the relative proportions of each of the chromosomal groups in the test dataset is similar to the relative proportions of each of the chromosomal groups in the training dataset.
17 . A cloud based computational system comprising one or more computing apparatus, comprising one or more processors and one or more memories wherein the cloud based computational system is configured to computationally generate an aneuploidy screening Artificial Intelligence (AI) model by:
defining a plurality of chromosomal group labels, wherein each group comprises one or more distinct aneuploidies comprising distinct genetic alterations or chromosomal abnormalities;
generating a training dataset from a first set of images, wherein each image comprises an image of an embryo captured after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present in at least one cell of the embryo, and the training dataset includes images labelled with each of the chromosomal groups;
generating a test dataset from a second set of images, wherein each image comprises an image of an embryo taken after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present, and the test dataset includes images labelled with each of the chromosomal groups;
separately train at least one chromosomal group AI model for each chromosomal group using the training dataset for training all models, wherein each chromosomal group AI model is trained to identify morphological features in images labelled with the associated chromosomal group label, and/or training at least one multi-group AI model on the training data wherein each multi-group AI model is trained to independently identify morphological features in images labelled with each of the associated chromosomal group labels to generate a multi-group output on an input image to indicate whether at least one aneuploidy associated with each of the chromosomal group is present or absent in the image;
selecting a best chromosomal group AI model for each of the chromosomal groups or a best multi-group AI model using the test dataset, wherein the aneuploidy screening AI model comprises the selected best chromosomal group AI model for each of the chromosomal groups or the best multi-group AI model, and in use, the aneuploidy screening AI model is used to screen embryo images for the presence of one or more aneuploidies.
18 . A cloud based computational system configured to computationally generate an estimate of the presence of one or more aneuploidies in an image of an embryo, wherein the computational system comprises:
one or more computation servers comprising one or more processors and one or more memories configured to:
store an aneuploidy screening Artificial Intelligence (AI) model configured to generate an estimate of the presence of one or more aneuploidies in an image of an embryo;
receive, from a user via a user interface of the computational system, an image;
provide the image to the aneuploidy screening AI model to obtain an estimate of the presence of one or more aneuploidies in the image; and
send a report on the presence of one or more aneuploidies in the image to the user via the user interface,
wherein the aneuploidy screening AI model is generated by:
defining a plurality of chromosomal group labels, wherein each group comprises one or more distinct aneuploidies comprising distinct genetic alterations or chromosomal abnormalities;
generating a training dataset from a first set of images, wherein each image comprises an image of an embryo captured after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present in at least one cell of the embryo, and the training dataset includes images labelled with each of the chromosomal groups;
generating a test dataset from a second set of images, wherein each image comprises an image of an embryo taken after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present, and the test dataset includes images labelled with each of the chromosomal groups;
separately train at least one chromosomal group AI model for each chromosomal group using the training dataset for training all models, wherein each chromosomal group AI model is trained to identify morphological features in images labelled with the associated chromosomal group label, and/or training at least one multi-group AI model on the training data wherein each multi-group AI model is trained to independently identify morphological features in images labelled with each of the associated chromosomal group labels to generate a multi-group output on an input image to indicate whether at least one aneuploidy associated with each of the chromosomal group is present or absent in the image;
selecting a best chromosomal group AI model for each of the chromosomal groups or a best multi-group AI model using the test dataset, wherein the aneuploidy screening AI model comprises the selected best chromosomal group AI model for each of the chromosomal groups or the best multi-group AI model.
19 . A computational system configured to generate an estimate of the presence of one or more aneuploidies in an image of an embryo, wherein the computational system comprises at least one processor, and at least one memory comprising instructions to configure the at least one processor to:
receive an image captured during a pre-determined time window after In-Vitro Fertilization (IVF);
upload, via a user interface, the image captured during a pre-determined time window after In-Vitro Fertilization (IVF) to a cloud based aneuploidy screening Artificial Intelligence (AI) model configured to generate an estimate of the presence of one or more aneuploidies in an image of an embryo;
receive an estimate of the presence of one or more aneuploidies in an image of an embryo via the user interface; and
display the estimate of the presence of one or more aneuploidies in an image of an embryo via the user interface,
wherein the cloud based aneuploidy screening AI model is generated by:
defining a plurality of chromosomal group labels, wherein each group comprises one or more distinct aneuploidies comprising distinct genetic alterations or chromosomal abnormalities;
generating a training dataset from a first set of images, wherein each image comprises an image of an embryo captured after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present in at least one cell of the embryo, and the training dataset includes images labelled with each of the chromosomal groups;
generating a test dataset from a second set of images, wherein each image comprises an image of an embryo taken after in-vitro fertilization, and is labelled with one or more chromosomal group labels where each label indicates if at least one aneuploidy associated with the respective chromosomal group is present, and the test dataset includes images labelled with each of the chromosomal groups;
separately train at least one chromosomal group AI model for each chromosomal group using the training dataset for training all models, wherein each chromosomal group AI model is trained to identify morphological features in images labelled with the associated chromosomal group label, and/or training at least one multi-group AI model on the training data wherein each multi-group AI model is trained to independently identify morphological features in images labelled with each of the associated chromosomal group labels to generate a multi-group output on an input image to indicate whether at least one aneuploidy associated with each of the chromosomal group is present or absent in the image;
selecting a best chromosomal group AI model for each of the chromosomal groups or a best multi-group AI model using the test dataset, wherein the aneuploidy screening AI model comprises the selected best chromosomal group AI model for each of the chromosomal groups or the best multi-group AI model.