Method for filtering normal medical image, method for interpreting medical image, and computing device implementing the methods

Inventors

Park, Jongchan

Assignees

Lunit Inc

Interested in licensing this patent?

MTEC can help explore whether this patent might be available for licensing for your application.

Publication Number

US-11928817-B2

Patent

Publication Date

2024-03-12

Expiration Date


Abstract

A method of reading a medical image by a computing device operated by at least one processor is provided. The method includes obtaining an abnormality score of the input image using an abnormality prediction model, filtering the input image so as not to be subsequently analyzed when the abnormality score is less than or equal to a cut-off score based on the cut-off score which makes a specific reading sensitivity; and obtaining an analysis result of the input image using a classification model that distinguishes the input image into classification classes when the abnormality score is greater than the cut-off score.

Core Innovation

The invention relates to a method of reading a medical image by a computing device that uses artificial intelligence to compute an abnormality score of an input image. A first stage analysis is performed based on the abnormality score, and the input image is classified as strong normal or non-strong normal. When the abnormality score is less than or equal to a cut-off score, the input image is classified as strong normal and is excluded from being subsequently analyzed by an artificial intelligence model that distinguishes the input image into classes including weak normal or abnormal.

After the first stage analysis, a second stage analysis is performed subsequent to the first stage analysis for input images having an abnormality score greater than the cut-off score. In the second stage analysis, an artificial intelligence model obtains an analysis result by distinguishing the input image into classes including weak normal or abnormal when the abnormality score is greater than the cut-off score. The two-stage approach separates strong normal from non-strong normal using the cut-off score and limits the later weak normal/abnormal classification to the remaining images.

The described approach includes selecting or determining the cut-off score based on reading sensitivity and a sensitivity/specificity trade-off, including a cut-off score tied to a specific reading sensitivity. The description further includes calibration of disease prediction scores using disease prediction head models, including per-disease cut-off scores and maximum across calibrated disease scores. The invention also describes feature extraction plus single-head or multi-head disease prediction architectures and uses a worklist and reporting outputs such as separate report, heatmaps, and secondary capture images for abnormal lesion visualization.

Claims Coverage

Two independent claims are provided, covering a method and a computing device for two-stage medical image reading based on an abnormality score and a cut-off score.

Two-stage medical image reading with abnormality score cut-off

obtains an abnormality score of an input image using an artificial intelligence model; performs a first stage analysis on the input image based on the abnormality score, performing the first stage analysis including classifying the input image as strong normal or non-strong normal, and excluding the input image classified as the strong normal from being subsequently analyzed by the artificial intelligence model that distinguishes the input image into classes including weak normal or abnormal when the abnormality score is less than or equal to a cut-off score; and performs a second stage analysis subsequent to the first stage analysis on the input image when the abnormality score is greater than the cut-off score, performing the second stage analysis including obtaining an analysis result of the input image using the artificial intelligence model that distinguishes the input image into classes including weak normal or abnormal when the abnormality score is greater than the cut-off score.

Computing device for two-stage analysis using abnormality score cut-off

obtains an abnormality score of an input image using an artificial intelligence model; performs a first stage analysis on the input image based on the abnormality score, performing the first stage analysis including classifying the input image as strong normal or non-strong normal, and excluding the input image classified as the strong normal from being subsequently analyzed by the artificial intelligence model that distinguishes the input image into classes including weak normal or abnormal when the abnormality score is less than or equal to a cut-off score; and performs a second stage analysis subsequent to the first stage analysis on the input image when the abnormality score is greater than the cut-off score, performing the second stage analysis including obtaining an analysis result of the input image using the artificial intelligence model that distinguishes the input image into classes including weak normal or abnormal when the abnormality score is greater than the cut-off score.

The independent claims center on a two-stage medical image reading workflow that computes an abnormality score, classifies strong normal versus non-strong normal using a cut-off score, excludes strong normal from subsequent weak normal/abnormal classification, and runs the second stage only when the abnormality score is greater than the cut-off.

Stated Advantages

Improves efficiency/workload by excluding strong normal images from subsequent analysis.

Supports reading sensitivity control by tying the cut-off score to desired sensitivity and using a sensitivity/specificity trade-off.

Documented Applications

Medical image reading for chest X-ray, including consolidation, nodule, and pneumothorax.

Visualization and reporting outputs using a worklist and separate report, including heatmaps and secondary capture images showing abnormal lesion position or predicted value.

JOIN OUR MAILING LIST

Stay Connected with MTEC

Keep up with active and upcoming solicitations, MTEC news and other valuable information.