Method and device for evaluating quality of pathological slide image
Inventors
PARK, Ga Hee • Paeng, Kyung Hyun • OCK, Chan Young • Song, Sang Hoon • Kim, Suk Jun
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
A computing device includes at least one memory, and at least one processor configured to analyze at least one object expressed in a pathological slide image, evaluate quality of the pathological slide image based on a result of the analyzing, and perform at least one additional operation according to a result of the evaluating.
Core Innovation
A computing device uses a machine learning model to identify an analyzable region and a cancer region from a pathological slide image, and generates first quantitative information corresponding to an area of the analyzable region and second quantitative information corresponding to an area of the cancer region. The system also determines quantitative information related to a biomarker based on at least one of a number of tumor cells, a number of lymphocyte cells, or phenotypes of tissues.
The computing device performs a first quality evaluation test by comparing the first quantitative information with a first reference value and by comparing the second quantitative information with a second reference value. Based on a result of the first quality evaluation test, the computing device performs a second quality evaluation test by comparing third quantitative information of the biomarker with a third reference value.
Based on a result of at least one of the first quality evaluation test or the second quality evaluation test, the computing device determines whether to perform downstream analysis of the pathological slide image. The system outputs a quality status via selective display and provides a report including evaluation results and a basis supporting those results.
Claims Coverage
The provided partial set includes two independent claims and multiple dependent refinements. The inventive features below summarize the main claim elements driving quality-evaluation gating of downstream analysis based on region area metrics and biomarker-related quantitative information compared to reference values.
Machine learning identification and region area quantification
Using a machine learning model to identify an analyzable region and a cancer region from a pathological slide image and generating first quantitative information corresponding to an area of the analyzable region and second quantitative information corresponding to an area of the cancer region.
First quality evaluation test using reference value comparison
Performing a first quality evaluation test of the pathological slide image based on a comparison between the first quantitative information and a first reference value and a comparison between the second quantitative information and a second reference value.
Biomarker quality evaluation test based on reference value comparison
Performing a second quality evaluation test by comparing third quantitative information of a biomarker with a third reference value, based on a result of the first quality evaluation test, wherein the third quantitative information of the biomarker is related to at least one of a number of tumor cells, a number of lymphocyte cells, or phenotypes of tissues.
Downstream analysis determination based on quality evaluation results
Determining, based on a result of at least one of the first quality evaluation test or the second quality evaluation test, whether to perform downstream analysis of the pathological slide image.
Region and cell classification into cancer-related and cell-type categories
Classifying regions of a pathological slide image into cancer-related and background categories and classifying cells within the image into tumor, lymphocyte, or other cell types.
Quality evaluation reporting with a basis for results
Generating a report that includes the result of the first quality evaluation test and a basis supporting that result.
Across the independent claims, the core coverage centers on machine learning identification of an analyzable region and a cancer region, generating quantitative area information and testing it against reference values, testing biomarker-related quantitative information against a reference value based on the first test result, and determining whether to perform downstream analysis based on the quality evaluation outcomes. Dependent features further specify region and cell classification, threshold-based pass/fail logic, and reporting that includes a basis for the evaluation result.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?