Apparatus, method, and system for image-based human embryo cell classification
Apparatuses, methods, and systems for automated cell classification, embryo ranking, and/or embryo categorization are provided. An apparatus includes a classification module configured to apply classifiers to images of one or more cells to determine, for each image, a classification probability associated with each classifier. Each classifier is associated with a distinct first number of cells, and is configured to determine the classification probability for each image based on cell features including one or more machine learned cell features. The classification probability indicates an estimated likelihood that the distinct first number of cells is shown in each image. The classification module is further configured to classify each image as showing a second number of cells based on the distinct first number of cells and the classification probabilities associated therewith. The classification module is implemented in at least one of a memory or a processing device.
1. A method for automated cell classification, comprising:
acquiring a series of time-sequential images of at least one human embryo comprising one or more cells contained in a multi-well culture dish comprising a plurality of microwells with a camera of at least one time-lapse microscope;
extracting image-based features from the series of time-sequential images;
applying a model to the image-based features;
applying a plurality of first classifiers to each of the series of time-sequential images of at least one human embryo to determine, for each image, a first classification probability associated with each first classifier, wherein:
each first classifier is associated with a distinct first number of cells, and determines the first classification probability for the each image based on a plurality of cell features including one or more machine learned cell features; and
the first classification probability indicates a first estimated likelihood that the distinct first number of cells associated with the each first classifier is shown in the each image, the each of the series of time-sequential images thereby having a plurality of the first classification probabilities associated therewith; and
classifying each image as showing a second number of cells based on the distinct first number of cells associated with the each first classifier and the plurality of first classification probabilities associated therewith, wherein the classification of each image as showing a second number of cells comprises one of the extracted image-based features;
generating an aneuploidy prediction for the at least one human embryo with the model and the image-based features.
2. The method of claim 1 , wherein the aneuploidy prediction comprises a categorization of the human embryo into one of several groups based on a score.
3. The method of claim 1 , further comprising providing a recommendation based on the aneuploidy prediction.
4. The method of claim 3 , wherein the recommendation is provided via a display.
5. The method of claim 3 , wherein the recommendation comprises one of: a selection of the human embryo for implantation; or a rejection of the human embryo.
6. The method of claim 5 , wherein the model comprises at least one of: a classifier; or a neural network.
7. The method of claim 6 , wherein the model is generated via machine learning.
8. A system for non-invasively predicting aneuploidy of a human embryo from a donor with an imaging system, the system comprising:
a stage configured to receive a multi-well culture dish comprising a plurality of micro-wells containing a sample comprising at least one human embryo;
a time-lapse microscope including a camera, a memory device, and a processor configured to acquire time-sequential images of the multi-well culture dish on the stage, wherein the time-lapse microscope is configured to:
extract image-based features from the series of time-sequential images and apply a model to the image-based features;
apply a plurality of first classifiers to each of the series of time-sequential images of at least one human embryo to determine, for each image, a first classification probability associated with each first classifier, wherein:
each first classifier is associated with a distinct first number of cells, and determines the first classification probability for the each image based on a plurality of cell features including one or more machine learned cell features; and
the first classification probability indicates a first estimated likelihood that the distinct first number of cells associated with the each first classifier is shown in the each image, the each of the plurality of images thereby having a plurality of the first classification probabilities associated therewith; and
classify each image as showing a second number of cells based on the distinct first number of cells associated with the each first classifier and the plurality of first classification probabilities associated therewith, wherein the classification of each image as showing a second number of cells comprises one of the extracted image-based features;
generate an aneuploidy prediction for the at least one human embryo with the model and the image-based features, wherein generating the aneuploidy prediction comprises generating a score associated with the development potential of the human embryo.
9. The system of claim 8 , wherein the score associated with the developmental potential of the human embryo is relative with respect to a plurality of other human embryos from the donor.
10. The system of claim 9 , wherein the aneuploidy prediction comprises a categorization of the human embryo into one of several groups based on the score.
11. The system of claim 8 , wherein the time-lapse microscope is further configured to generate a recommendation based on the aneuploidy prediction.
12. The system of claim 11 , further comprising a display configured to provide a recommendation based on the aneuploidy prediction.
13. The system of claim 12 , wherein the recommendation comprises one of: a selection of the human embryo for implantation; or a rejection of the human embryo.
14. The system of claim 13 , wherein the model comprises at least one of: a classifier; or a neural network.