Method and system for predicting expression of biomarker from medical image
Inventors
AUM, Jae Hong • OCK, Chanyoung • Yoo, Donggeun
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12136218-B2
Publication Date
2024-11-05
Expiration Date
Abstract
The present disclosure relates to a method for predicting biomarker expression from a medical image. The method for predicting biomarker expression includes receiving a medical image, and outputting indices of biomarker expression for the at least one lesion included in the medical image by using a first machine learning model.
Core Innovation
A method obtains a medical image created by capturing at least one part of a patient's body without tissue collection, inputs the medical image to a machine learning model, and outputs information related to a biomarker expression for each of a plurality of lesions detected in the medical image. The method determines information associated with tissue collection for the plurality of lesions in the medical image based on the information related to the biomarker expression.
Lesion regions are extracted and cropped from the medical image to generate partial images. The partial images can be used as inputs to the machine learning model for outputting information related to the biomarker expression, and segmentation information can be determined for each lesion and provided with the partial images as input to the machine learning model.
The approach further supports multi-timepoint imaging by obtaining a first and second medical image of a patient's body at different time points. Lesion regions are extracted from both images and input to the machine learning model to output information related to the biomarker expression, and biomarker-expression information associated with a different lesion can be used to improve prediction.
Claims Coverage
The document contains two independent claims covering a core pipeline that predicts biomarker expression for detected lesions from medical images obtained without tissue collection, and then determines tissue-collection-associated information based on the predicted biomarker-expression information. Across the dependent claims, the coverage adds lesion-region processing, optional segmentation inputs, multi-timepoint imaging inputs, and tissue-collection attributes including tissue-collection priority determined using reference tissue-collection information. In total, the independent claim set covers 10 inventive-feature concepts across the main independent logic plus its main dependent refinements.
Predict biomarker expression without tissue collection
Obtain a medical image created by capturing at least one part of a patient's body without tissue collection, input the medical image to a machine learning model, and output information related to a biomarker expression for each of a plurality of lesions in the medical image.
Determine tissue collection information based on biomarker expression
Determine information associated with tissue collection for the plurality of lesions in the medical image based on the information related to the biomarker expression.
Extract lesion regions and crop to partial images
Extract regions containing at least one of the plurality of lesions from the medical image and crop the extracted regions to generate partial images.
Use segmentation information with partial images
Determine segmentation information for each lesion, and input the segmentation information and partial images to the machine learning model to output the information related to the biomarker expression.
Multi-timepoint lesion-region input for biomarker prediction
Obtain a first and second medical image of a patient's body at different time points, extract lesion regions from both images, and input the extracted regions into the machine learning model to output information related to the biomarker expression.
Tissue collection attributes include method, location, and/or priority
The information associated with the tissue collection comprises at least one of the tissue collection method, the tissue collection location, or the tissue collection priority.
Determine tissue collection priorities using reference information
Acquire reference information about the tissue collection associated with a medical image, and determine tissue-collection priority information for each of a plurality of lesions by determining the priorities using the biomarker-expression information together with the reference information.
Across the two independent claims, the invention is directed to predicting biomarker-expression information for each detected lesion from a medical image obtained without tissue collection using a machine learning model, and then determining tissue-collection-associated information for the lesions based on the predicted biomarker expression. Dependent claim coverage further specifies lesion-region extraction and cropping to partial images, optional use of segmentation information, support for multi-timepoint imaging inputs, and tissue-collection information that can include method/location/priority, including determining tissue-collection priorities using reference information.
Stated Advantages
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?