System and method for tumor characterization
Inventors
Wang, Jing • Kulkarni, Prathamesh • Robinson, Eric
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12536223-B2
Publication Date
2026-01-27
Expiration Date
Abstract
A method of treating a subject comprises administering a treatment to a subject identified as having a high probability of distant metastatic recurrence, wherein the probability of distant metastatic recurrence was determined by a process, comprising acquiring at least one image of a tissue sample comprising a plurality of cells, taken from a subject, classifying each of the plurality of cells into categories, dividing the at least one image into a plurality of patches, calculating values for a plurality of morphological features based on the patches, and calculating a distant metastatic recurrence probability based on the values. A computer-implemented method of training a neural network and a system for characterizing a cancer in a subject are also described.
Core Innovation
The invention describes a computer-implemented method of training a neural network for characterizing a cancer in a body of a subject. The method acquires at least one first image of a first tissue sample including a plurality of cells, classifies each cell into categories, and divides the image into a plurality of patches in which each patch comprises multiple cells. The method then calculates a cell density for each patch and calculates values for a plurality of morphological features based on the patches.
The morphological features are expressed as ratios or area-based relationships involving counts and totals of immune cells in large cluster, immune cells in small cluster, tumor cells, and combined tumor+immune quantities, including immune cell total area relative to tumor cell total area and relative to a combined immune area+tumor area. A distant metastatic recurrence probability is calculated based on the morphological feature values and the cell density.
The neural network is trained using a combination of the morphological feature values and a low-dimensional representation of a sequence generated by a DNN. In the provided claim sets, random selection is used to divide at least one image into patches, and patches having no tumor cell information from the plurality of patches are discarded.
Claims Coverage
The partial content includes two independent claims. Both share the same core set of inventive features: cell classification in tissue images, patching, cell-density calculation, morphological feature value calculation, and training a neural network using those feature values together with a low-dimensional representation of a sequence generated by a DNN, with a distant metastatic recurrence probability as an output for the overall characterization.
Training neural network for distant metastatic recurrence from tissue-image patch morphology and sequence representation
Acquiring at least one first image of a first tissue sample comprising a plurality of cells; classifying each of the plurality of cells into categories; dividing the at least one image into a plurality of patches; calculating a cell density of each patch; calculating values for a plurality of morphological features based on the patches; calculating a distant metastatic recurrence probability based on the values and the cell density; and training at least one neural network using a combination of the values and a low-dimensional representation of a sequence generated by a DNN.
Training neural network with random patching and discarded non-informative patches using patch morphological features and sequence representation
Acquiring at least one image of a tissue sample comprising a plurality of cells; classifying each cell into categories; dividing the at least one image into a plurality of patches at least partially by random selection; discarding patches having no tumor cell information from the plurality of patches; calculating values for a plurality of morphological features based on the patches; and training at least one neural network using a combination of the values and a low-dimensional representation of a sequence generated by a DNN, wherein the morphological features comprise at least one of the specified immune- and tumor-related count and area ratios.
Across the two independent claims, the inventive scope centers on computing patch-level cell density and predefined morphological feature ratios from classified tissue-image cells, then training a neural network using those feature values together with a low-dimensional sequence representation generated by a DNN to characterize cancer via a distant metastatic recurrence probability.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?