| CPC G06V 10/774 (2022.01) [G06T 7/0014 (2013.01); G06V 10/82 (2022.01); G06V 2201/03 (2022.01)] | 20 Claims |
|
1. A system, comprising:
a memory configured to store computer-executable components; and
a processor that executes at least one of the computer-executable components that:
trains, using a set of training data, a machine learning model to perform a regression task on medical images, wherein the set of training data comprises tuples,
wherein each tuple comprises:
a respective training medical image, wherein the set of training data comprises a plurality of different training medical images,
a respective training confidence quantile value for the respective training medical image, wherein the set of training data comprises a plurality of different training confidence quantile values, and
a respective ground truth annotation;
wherein the machine learning model comprises:
a set of hidden layers,
an input layer coupled to the set of hidden layers, wherein the input layer provides first inputs to the set of hidden layers,
a conditioning layer coupled to the set of hidden layers, wherein the input layer is parallel to the conditioning layer, and wherein the conditioning layer provides second inputs to the set of hidden layers, and
an output layer that receives outputs from the set of hidden layers; and
wherein the training comprises iteratively performing until a defined criterion is satisfied:
selecting a tuple from the set of training data that has not been employed for the training,
providing the respective training medical image of the tuple to the input layer,
providing the respective training confidence quantile value of the tuple to the conditioning layer,
performing, via execution of the machine learning model, the regression task on the respective training medical image using the respective training confidence quantile value, and
updating the machine learning model based on loss function associated with a result of performing the regression task and the respective ground truth annotation of the tuple;
accesses a medical image and a user-specified confidence quantile value; and
performs, via execution of the machine learning model, the regression task on the medical image with the user-specified confidence quantile value.
|