US 12,272,063 B2
Apparatuses and methods for training and using computational operations for digital image processing
Benoit Huet, Roquefort les Pins (FR); Pierre Baudot, Marseilles (FR); Elias Munoz, Biot (FR); Ezequiel Geremia, Mougins (FR); Jean-Christophe Brisset, Valbonne-Sophia Antipolis (FR); and Vladimir Groza, Novosibirsk (RU)
Assigned to Median Technologies, Valbonne (FR)
Filed by Median Technologies, Valbonne (FR)
Filed on Aug. 16, 2024, as Appl. No. 18/807,820.
Application 18/807,820 is a continuation of application No. PCT/EP2023/053233, filed on Feb. 9, 2023.
Prior Publication US 2024/0412363 A1, Dec. 12, 2024
Int. Cl. G06T 7/00 (2017.01); G06T 7/11 (2017.01); G06T 7/62 (2017.01); G06V 10/26 (2022.01); G06V 10/764 (2022.01); G06V 10/77 (2022.01); G06V 10/774 (2022.01)
CPC G06T 7/0012 (2013.01) [G06T 7/11 (2017.01); G06T 7/62 (2017.01); G06V 10/26 (2022.01); G06V 10/764 (2022.01); G06V 10/7715 (2022.01); G06V 10/774 (2022.01); G06T 2200/04 (2013.01); G06T 2207/20036 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/30004 (2013.01); G06V 2201/03 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method for training at least one computing operation using 3-dimensional images and using at least one trained computing operation on an input 3-dimensional image, the method comprising:
receiving, from a first geographical location, at a first time, using one or more hardware computing device processors, a first image comprising a first 3-dimensional image displaying at least one detectable structure, wherein the at least one detectable structure comprises at least one quantifiable visual feature;
determining, using the one or more hardware computing device processors, based on the first image, the at least one detectable structure;
segmenting, using the one or more hardware computing device processors, the first image to obtain a first segmentation mask, wherein the first segmentation mask is associated with at least one geometric shape and comprises the at least one quantifiable visual feature;
generating, using the one or more hardware computing device processors, based on the first segmentation mask, a first mesh based on or comprising the at least one quantifiable visual feature,
wherein the first mesh is associated with a first surface,
wherein the first surface is associated with a first set of vertices;
geometrizing the first mesh, based on a first geometric parameter, thereby resulting in a first geometrized mesh;
computing or determining, using the one or more hardware computing device processors, at least one first quantifiable visual parameter based on the first mesh;
extracting, using the one or more hardware computing device processors, a first quantifiable visual data from the first image based on the at least one first quantifiable visual parameter;
training the at least one computing operation using the first quantifiable visual data as a first input;
receiving, from the first geographical location or a second geographical location, at the first time or a second time, using the one or more hardware computing device processors, a second image comprising a second 3-dimensional image displaying the at least one detectable structure, wherein the at least one detectable structure comprises the at least one quantifiable visual feature;
determining, using the one or more hardware computing device processors, based on the second image, the at least one detectable structure;
segmenting, using the one or more hardware computing device processors, the second image to obtain a second segmentation mask, wherein the second segmentation mask is associated with the at least one geometric shape and comprises the at least one quantifiable visual feature;
generating, using the one or more hardware computing device processors, based on the second segmentation mask, a second mesh based on or comprising the at least one quantifiable visual feature,
wherein the second mesh is associated with a second surface,
wherein the second surface is associated with a second set of vertices;
geometrizing the second mesh, based on the first geometric parameter or a second geometric parameter, thereby resulting in a second geometrized mesh;
computing or determining, using the one or more hardware computing device processors, at least one second quantifiable visual parameter for the second mesh;
extracting, using the one or more hardware computing device processors, a second quantifiable visual data from the second image based on the at least one second quantifiable visual parameter;
training the at least one computing operation using the second quantifiable visual data as a second input;
receiving, from the first geographical location, the second geographical location, or a third geographical location, at the first time, the second time, or a third time, using the one or more hardware computing device processors, a third image comprising a third 3-dimensional image displaying the at least one detectable structure, wherein the at least one detectable structure comprises the at least one quantifiable visual feature;
determining, using the one or more hardware computing device processors, based on the third image, the at least one detectable structure;
segmenting, using the one or more hardware computing device processors, the third image to obtain a third segmentation mask, wherein the third segmentation mask is associated with the at least one geometric shape and comprises the at least one quantifiable visual feature;
generating, using the one or more hardware computing device processors, based on the third segmentation mask, a third mesh based on or comprising the at least one quantifiable visual feature,
wherein the third mesh is associated with a third surface,
wherein the third surface is associated with a third set of vertices;
geometrizing the third mesh, based on the first geometric parameter, the second geometric parameter, or a third geometric parameter, thereby resulting in a third geometrized mesh;
computing or determining, using the one or more hardware computing device processors, at least one third quantifiable visual parameter for the third mesh;
extracting, using the one or more hardware computing device processors, a third quantifiable visual data from the third image based on the at least one third quantifiable visual parameter;
training the at least one computing operation using the third quantifiable visual data as a third input;
receiving, from a fourth geographical location that is different from all of the first geographical location, the second geographical location, and the third geographical location, at a fourth time that is different from all of the first time, the second time, and the third time, using the one or more hardware computing device processors, a fourth image comprising a fourth 3-dimensional image displaying the at least one detectable structure;
determining, using the one or more hardware computing device processors, based on the fourth image, the at least one detectable structure;
segmenting, using the one or more hardware computing device processors, the fourth image to obtain a fourth segmentation mask, wherein the fourth segmentation mask is associated with the at least one geometric shape and comprises the at least one quantifiable visual feature;
generating, using the one or more hardware computing device processors, based on the fourth segmentation mask, a fourth mesh comprising the at least one quantifiable visual feature,
wherein the fourth mesh is associated with a fourth surface,
wherein the fourth surface is associated with a fourth set of vertices;
geometrizing the fourth mesh, based on the first geometric parameter, the second geometric parameter, the third geometric parameter, or a fourth geometric parameter, thereby resulting in a fourth geometrized mesh;
computing or determining, using the one or more hardware computing device processors, at least one fourth quantifiable visual parameter for the fourth mesh;
extracting, using the one or more hardware computing device processors, a fourth quantifiable visual data from the fourth image based on the at least one fourth quantifiable visual parameter;
classifying, based on the training the at least one computing operation using the first quantifiable visual data as the first input, the training the at least one computing operation using the second quantifiable visual data as the second input, and the training the at least one computing operation using the third quantifiable visual data as the third input, using the one or more hardware computing device processors, the fourth quantifiable visual data, thereby resulting in a classification;
generating, using the one or more hardware computing device processors, based on the classification, a computation result for the fourth image; and
correlating, using the one or more hardware computing device processors, the computation result with a malignancy dataset associated with the at least one detectable structure, thereby resulting in correlation data.