Generating final abnormality data for medical scans based on utilizing a set of sub-models

Patent metadata

Field	Value
Publication number	US-11922348-B2
Application number	US-202217656925-A
Country	US
Kind code	B2
Filing date	Mar 29, 2022
Priority date	Nov 21, 2018
Publication date	Mar 5, 2024
Grant date	Mar 5, 2024

Abstract

A multi-model medical scan analysis system is operable to generate a plurality of training sets from a plurality of medical scans. Each of a set of sub-models is generated by performing a training step on a corresponding one of the plurality of training sets of the plurality of medical scans. A set of abnormality data is generated by applying a subset of a set of inference functions on a new medical scan. The subset of the set of inference functions utilize the subset of the set of sub-models, and each of the set of abnormality data is generated as output of performing one of the subset of the set of inference functions. The multi-model medical scan analysis system is further operable to generate final abnormality data that includes a global probability indicating a probability that any abnormality is present based on the set of abnormality data.

First claim

What is claimed is:

1. A multi-model medical scan analysis system, comprising:
at least one processor; and
a memory that stores operational instructions that, when executed by the at least one processor, cause the multi-model medical scan analysis system to:
generate a plurality of training sets from a plurality of medical scans;
generate each of a set of sub-models by performing a training step on a corresponding one of the plurality of training sets of the plurality of medical scans;
receive a new medical scan;
generate a set of abnormality data by applying a subset of a set of inference functions on the new medical scan, wherein the subset of the set of inference functions utilize a subset of the set of sub-models, wherein each of the set of abnormality data is generated as output of performing one of the subset of the set of inference functions;
generate final abnormality data that includes a global probability indicating a probability that any abnormality is present based on the set of abnormality data; and
transmit the final abnormality data for display via a display device of a client device.

2. The multi-model medical scan analysis system of claim 1 , wherein the set of abnormality data indicates a plurality of probabilities of abnormality detection, and wherein the global probability is based on the plurality of probabilities of abnormality detection.

3. The multi-model medical scan analysis system of claim 2 , wherein a first one of the set of sub-models is trained to detect a first type of abnormality, wherein a second one of the set of sub-models is trained to detect a second type of abnormality, wherein a first one of the set of abnormality data is generated as output of a first one of the subset of the set of inference functions that corresponds to the first one of the set of sub-models, wherein a second one of the set of abnormality data is generated as output of a second one of the subset of the set of inference functions that corresponds to the second one of the set of sub-models, wherein the first one of the set of abnormality data indicates a first probability that the first type of abnormality is present, wherein the second one of the set of abnormality data indicates a second probability that second type of abnormality is present, and wherein the plurality of probabilities of abnormality detection includes the first probability and the second probability.

4. The multi-model medical scan analysis system of claim 1 , wherein the operational instructions when executed by the at least one processor, cause the multi-model medical scan analysis system to select a subset of the set of sub-models based on the new medical scan.

5. The multi-model medical scan analysis system of claim 4 , wherein at least two of the set of sub-models each correspond to different ones of a set of medical scan classification categories, wherein at least two of the plurality of training sets are generated to include ones of the plurality of medical scans of a corresponding one of the set of medical scan classification categories, and wherein a subset of the set of sub-models are selected by determining a corresponding subset of the set of medical scan classification categories that compare favorably to the new medical scan.

6. The multi-model medical scan analysis system of claim 5 , wherein the new medical scan is received in a study for a patient that includes a plurality of medical scans that each include a plurality of image slices, wherein a subset of the set of sub-models is selected based on the plurality of medical scans in the study, and wherein at least two of the subset of the set of inference functions are performed on different ones of the plurality of medical scans to generate the set of abnormality data.

7. The multi-model medical scan analysis system of claim 6 , wherein the set, of abnormality data indicates a plurality of probabilities of abnormality detection, wherein the plurality of probabilities of abnormality detection indicate whether an abnormality is present in the different ones of the plurality of medical scans, and wherein the final abnormality data indicates a probability that abnormality is present in the patient given the plurality of probabilities of the set of abnormality data based on utilizing a Bayesian model.

8. The multi-model medical scan analysis system of claim 1 , wherein generating the final abnormality data includes performing a final inference function.

9. The multi-model medical scan analysis system of claim 8 , wherein the final inference function utilizes a plurality of known correlations between different types of abnormalities, and wherein the final abnormality data is generated based on a known correlation between a first type of abnormality and a second type of abnormality.

10. The multi-model medical scan analysis system of claim 9 , wherein one of: the final abnormality data indicates an increase in a second probability that the second type of abnormality is present in response to the set of abnormality data indicating a first probability that the first type of abnormality is present comparing favorably to a detection threshold and in response to the known correlation between the first type of abnormality and the second type of abnormality comparing favorably to a correlation threshold; or the final abnormality data indicates a decrease in the second probability that the second type of abnormality is present in response to the set of abnormality data indicating the first probability that the first type of abnormality is present comparing unfavorably to a detection threshold and in response to the known correlation between the first type of abnormality and the second type of abnormality comparing favorably to a correlation threshold.

11. The multi-model medical scan analysis system of claim 1 , wherein the operational instructions, when executed by at least one processor, further cause the multi-model medical scan analysis system to:
partition image data of one of the plurality of medical scans into at least two partitioned image data portions in accordance with a plurality of partitioning categories, wherein a first one of the plurality of training sets includes a first partitioned image data portion of the least two partitioned image data portions, and wherein a second one of the plurality of training sets includes a second partitioned image data portion of the least two partitioned image data portions; and
partition image data of the new medical scan into at least two new partitioned image data portions, wherein the first one of the subset of the set of inference functions is performed upon a first new partitioned image data portion of at least two new partitioned image data portions, and wherein the second one of the subset of the set of inference functions is performed upon a second new partitioned image data portion of at least two new partitioned image data portions.

12. The multi-model medical scan analysis system of claim 11 , wherein partitioning the image data of the one of the plurality of medical scans into at least two partitioned image data portions includes one of: partitioning a set of images slices included in the one of the plurality of medical scans into different subsets of the set of image slices for inclusion in different ones of the plurality of training sets; or cropping at least one image slice included in the one of the plurality of medical scans into different cropped portions for inclusion in different ones of the plurality of training sets.

13. The multi-model medical scan analysis system of claim 11 , wherein image data of each of the plurality of medical scans is partitioned by a plurality

Assignees

• Enlitic Inc

Inventors

• Lyman Kevin
• Yao Li
• Poblenz Eric C
• Prosky Jordan
• Covington Ben
• Upton Anthony

Classifications

• G06T12/20

  Inverse problem, i.e. transformations from projection space into object space · CPC title

• G06T11/26

  Drawing of charts or graphs · CPC title

• G06T11/10

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/09

  Supervised learning · CPC title