Systems and methods for determining tissue microarray sampling protocols
Inventors
Woicik, Adelaide • Probert, Christopher • SERRANO, Santiago Akle • McCaw, Zachary R. • Dulken, Benjamin • NARAYANAN, Sanjana
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12333725-B2
Publication Date
2025-06-17
Expiration Date
Abstract
An exemplary method for determining a sampling protocol for sampling tissue cores for a tissue microarray includes obtaining an initial plurality of tissue cores from an image of a tissue slide; selecting a first subset of the initial plurality of tissue cores based on a first candidate sampling protocol; inputting the first subset of the plurality of tissue cores into a machine learning model; evaluating the first candidate sampling protocol by evaluating a first output of the machine learning model; selecting a second subset of the initial plurality of tissue cores based on a second candidate sampling protocol; inputting the second subset of the plurality of tissue cores into the machine learning model; evaluating the second candidate sampling protocol by evaluating a second output of the machine learning model; and determining the sampling protocol based on the evaluation of the first candidate sampling protocol and the second candidate sampling protocol.
Core Innovation
The invention determines a sampling protocol for sampling tissue cores for a tissue microarray by receiving an initial plurality of tissue cores from an image of a tissue slide. The system selects a first subset of the initial plurality of tissue cores based on a first candidate sampling protocol and inputs the first subset into a machine learning model. It evaluates the first candidate sampling protocol by evaluating a first output of the machine learning model based on the first subset of the plurality of tissue cores.
The system selects a second subset of the initial plurality of tissue cores based on a second candidate sampling protocol and inputs the second subset into the machine learning model. It evaluates the second candidate sampling protocol by evaluating a second output of the machine learning model based on the second subset of the plurality of tissue cores. The system determines the sampling protocol based on the evaluation of the first candidate sampling protocol and the second candidate sampling protocol.
In dependent refinements, the initial plurality of tissue cores is generated using integer linear programming by maximizing an objective function subject to a set of constraint parameters. The constraint parameters include tissue-core size and spacing or tissue-content and edge-distance requirements, and the candidate sampling protocol includes parameters defining tissue-core count, tissue-core size, minimum distance between tissue cores and an edge of a tissue, and one or more tissue labels for regions of the tissue slide.
The tissue labels are generated from tissue labels assigned to regions of the image by using an image segmentation model, and embeddings are generated from tissue cores in a subset and input to the machine learning model.
Claims Coverage
The document contains three independent claims: clm-00001 (system), clm-00018 (method), and clm-00029 (non-transitory computer-readable medium). Across these independent claims, the core claim coverage centers on evaluating multiple candidate sampling protocols by selecting different subsets of an initial plurality of tissue cores and using machine learning model outputs to determine a sampling protocol, with dependent claims refining how initial cores, subsets, protocol parameters, labels, and model inputs are produced and evaluated.
Evaluating candidate sampling protocols using machine learning model outputs
select a first subset of the initial plurality of tissue cores based on a first candidate sampling protocol; input the first subset of the plurality of tissue cores into a machine learning model; evaluate the first candidate sampling protocol by evaluating a first output of the machine learning model based on the first subset of the plurality of tissue cores; select a second subset of the initial plurality of tissue cores based on a second candidate sampling protocol; input the second subset of the plurality of tissue cores into the machine learning model; evaluate the second candidate sampling protocol by evaluating a second output of the machine learning model based on the second subset of the plurality of tissue cores; and determine the sampling protocol based on the evaluation of the first candidate sampling protocol and the second candidate sampling protocol.
Generating initial plurality of tissue cores using integer linear programming
receive an initial plurality of tissue cores from an image of a tissue slide; and obtain the initial plurality of tissue cores using integer linear programming by maximizing an objective function defined by a set of constraint parameters.
Parameterizing candidate sampling protocols with tissue and spatial constraints and tissue labels
select a first subset of the initial plurality of tissue cores based on a first candidate sampling protocol, wherein the first candidate sampling protocol includes one or more first tissue-core sampling parameters including a first number of the tissue cores, a first tissue core size, a first amount of tissue in each core of the first subset, a first minimum distance between each tissue core of the first subset and an edge of a tissue, and one or more first tissue labels for regions of the tissue slide.
Producing tissue labels and embeddings for machine learning model input
generate one or more first tissue labels from tissue labels assigned to regions of an image by providing the image to an image segmentation model; and generate one or more embeddings from each tissue core in a first subset and input the one or more embeddings into a machine learning model.
Overall, the claim set is directed to determining a tissue microarray sampling protocol by selecting different subsets of an initial plurality of tissue cores according to candidate sampling protocols, evaluating each candidate using machine learning model outputs, and determining the sampling protocol based on those evaluations. Dependent coverage further specifies generating the initial core set via integer linear programming with constraint parameters, parameterizing candidate protocols with tissue/spatial constraints and region tissue labels, generating those labels via an image segmentation model, and generating embeddings from tissue cores for machine learning model input.
Stated Advantages
Cost/time efficiency for generating sampling protocols.
Standardized protocol generation.
Reduced storage/compute compared to whole slide images.
Documented Applications
No documented applications found
Interested in licensing this patent?