Method of processing an image of tissue, a system for processing an image of tissue, a method for disease diagnosis and a disease diagnosis system
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-12175661-B2
Publication Date
2024-12-24
Expiration Date
Abstract
A computer implemented method of processing an image of tissue, comprising: inputting image data comprising a plurality of pixels into a first trained model, the first trained model generating a value corresponding to each of a plurality of pixels representing a feature relevant to disease diagnosis; wherein the first trained model comprises a convolutional neural network comprising a layer in which a first filter and a second filter are applied, at least one of the first filter and the second filter comprising a dilated convolution, wherein the output of the first filter and the second filter are combined and input into the subsequent layer.
Core Innovation
The invention provides a computer implemented method of processing an image of tissue for disease diagnosis. Image data corresponding to the image of tissue is input into a first trained model comprising a convolutional neural network and into a second trained model comprising a convolutional neural network, and the first set of output data and the second set of output data are combined.
Diagnostic information is generated from the combined data using a third trained model, and a diagnosis is generated from the diagnostic information and context information using a fourth trained model. The diagnostic information is based on disease-relevant features derived from the convolutional neural networks, and the diagnosis is based on both the diagnostic information and context information.
The description further characterizes the system as a tissue image analysis pipeline using a convolutional neural network for pixel-wise semantic image segmentation of disease-relevant features. The convolutional neural network includes dilated (atrous) convolution and skip connection architectures with multiple dilation factors to learn local and non-local morphology.
Claims Coverage
The document provides three independent claims: a computer implemented method for processing tissue images to generate diagnosis from combined model outputs with context, a computer implemented method of training the multi-model system, and a system for processing tissue images configured to perform the same multi-model workflow. Across the independent claims, the core inventive structure uses four trained models with two convolutional neural networks producing disease-relevant feature outputs that are combined, followed by models that produce diagnostic information and a final diagnosis using context information.
Two-model feature extraction from tissue image data and combining outputs
Inputting image data corresponding to the image of tissue into a first trained model comprising a convolutional neural network generating a first set of output data representing a feature relevant to disease diagnosis; inputting the image data into a second trained model comprising a convolutional neural network generating a second set of output data representing a feature relevant to disease diagnosis; combining the first set of output data and the second set of output data.
Diagnostic information generation from combined features
Generating diagnostic information from the combined data using a third trained model.
Diagnosis generation from diagnostic information and context information
Generating a diagnosis from the diagnostic information and context information using a fourth trained model.
Training a four-model tissue-image processing system using training data labels
Training the first model using training data labels; training the second model using training data labels; generating diagnostic information from training data using a third model; training the third model using training data labels; generating a diagnosis from the diagnostic information and context information using a fourth model; and training the fourth model using training data labels.
System configured with multi-model tissue image processing pipeline
A system for processing an image of tissue comprising an input for receiving image data; an output for outputting diagnostic information; and a processor configured to input the received image data into a first trained model and a second trained model comprising convolutional neural networks, combine their output data, generate diagnostic information from the combined data using a third trained model, and generate a diagnosis from the diagnostic information and context information using a fourth trained model.
Overall claim coverage centers on processing tissue image data through two convolutional neural network-based feature models, combining their output data, generating diagnostic information with a third trained model, and generating a diagnosis with a fourth trained model using diagnostic information and context information; additional claim coverage includes training each model using training data labels and providing a system implementation of the same multi-model pipeline.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?