Systems and methods for analyzing two-dimensional and three-dimensional image data
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12367574-B2
Publication Date
2025-07-22
Expiration Date
Abstract
The present disclosure provides a computerized method for a likelihood of malignancy in breast tissue of a patient. The method includes receiving, with a computer processor, an image of the breast tissue, providing the image of the breast tissue to a model including a trained neural network; the trained neural network being previously trained by training a first neural network, initializing a second neural network based on the first neural network, training the second neural network, and outputting the second neural network as the trained neural network, receiving an indicator from the model, and outputting a report including the indicator to at least one of a memory or a display.
Core Innovation
The invention relates to generating an indication and/or a malignancy likelihood score for breast tissue of a patient using a computerized method that receives an image of the breast tissue. In at least some embodiments, the image is a synthetic two-dimensional image generated from an image that comprises a plurality of images, each image being associated with a location value, based on a subset of the plurality of images associated with location values included in a predetermined range of location values.
The method provides the image of the breast tissue to a model comprising a trained neural network. The trained neural network is previously trained using a two-stage training process in which a first neural network is trained based on a set of annotated patches derived from a first group of two-dimensional images, with each annotated patch comprising a patch-level label and a portion of a two-dimensional image.
A second neural network is initialized based on the first neural network and is trained based on annotated images. The second neural network training includes training with a first set of annotated images comprising at least one two-dimensional image, at least one bounding box, and at least one malignancy label or malignancy likelihood score associated with the bounding box, and training with a second set of annotated images comprising at least one two-dimensional image and an image-level malignancy likelihood score. The second neural network is output as the trained neural network, and an indicator or a malignancy likelihood score is received from the model.
Claims Coverage
The independent claims cover three core aspects: selection of a subset of location-value-associated images to generate a synthetic two-dimensional image, two-stage training of a neural network using annotated patches and annotated images with bounding boxes and malignancy labels or scores, and outputting an indicator and/or a malignancy likelihood score from the model.
Synthetic two-dimensional image from subset of location values
Receiving an image of the breast tissue, wherein the image of the breast tissue is a synthetic two-dimensional image generated based on an image that comprises a plurality of images, each image included in the plurality of images being associated with a location value, and wherein the synthetic two-dimensional image is generated based on a subset of the plurality of images associated with location values included in a predetermined range of location values.
Two-stage neural network training from annotated patches to trained neural network
Providing the image of the breast tissue to a model comprising a trained neural network, the trained neural network being previously trained by training a first neural network based on a set of annotated patches derived from a first group of two-dimensional images, each annotated patch comprising a patch-level label and a portion of a two-dimensional image included in the first group of two-dimensional images, initializing a second neural network based on the first neural network, training the second neural network based on a first set of annotated images comprising at least one bounding box and at least one malignancy label associated with the at least one bounding box, and training the second neural network based on a second set of annotated images comprising a two-dimensional image and an image-level malignancy likelihood score, and outputting the second neural network as the trained neural network.
Indicator from trained neural network for likelihood of malignancy
Receiving an indicator from the model, wherein the indicator is an indication of a likelihood of malignancy in breast tissue of a patient.
Malignancy likelihood score output from trained neural network
Receiving the malignancy likelihood score for the breast tissue of the patient from the model.
System with memory and processor for indicator or malignancy likelihood score generation
A memory configured to store an image of the breast tissue and a processor configured to access the memory, provide the image to a model comprising a trained neural network trained by training a first neural network based on annotated patches, initializing a second neural network based on the first neural network, training the second neural network based on annotated images with at least one bounding box and at least one malignancy label associated with the at least one bounding box, and training the second neural network based on annotated images comprising an image-level malignancy likelihood score, and receive an indicator from the model or receive the malignancy likelihood score for the breast tissue of the patient from the model.
Across the independent claims, the inventive coverage centers on selecting a subset of location-value-associated images to generate a synthetic two-dimensional image of breast tissue, using a two-stage trained neural network based on annotated patches and annotated images with bounding boxes and image-level malignancy likelihood score, and outputting an indicator and/or a malignancy likelihood score from the trained model.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?