US 11,776,130 B2 · Progressively-trained scale-invariant and boundary-aware deep neural network for the automatic 3D segmentation of lung lesions

Abstract

A system and method are disclosed for segmenting a set of two-dimensional CT slices corresponding to a lesion. In an embodiment, for each of at least a subset of the set of CT slices, the system inputs the CT slice into a plurality of branches of a trained segmentation block. Each branch of the segmentation block includes a convolutional neural network (CNN) with filters at a different scale, and produces a plurality of levels of output. The system generates, for each CT slice in the subset, feature maps for each level of output. The system generates a segmentation of each CT slice in the subset based on the feature maps of each level of output. The system aggregates the segmentations of each slice in the subset to generate a three-dimensional segmentation of the lesion. The system provides data representing the three-dimensional segmentation for display.

Claims20

Dependent claims

A computer-implemented method of segmenting a set of two-dimensional computed tomography (CT) slices corresponding to a lesion, comprising:

for each of at least a subset of the set of CT slices:
inputting the CT slice into a plurality of branches of a trained segmentation block, wherein each branch includes a convolutional neural network (CNN) with convolutional filters at a different scale, and each branch produces a plurality of levels of output;
generating feature maps for each of the levels of outputs, the feature maps generated by combining same-level outputs across each of the branches;
and
generating a segmentation of the CT slice based on the feature maps;
aggregating the segmentations of each slice in the subset of CT slices to generate a three-dimensional segmentation of the lesion;
and
providing data representing the three-dimensional segmentation of the lesion for display of the three-dimensional segmentation.

5 more elements

A non-transitory computer-readable storage medium containing computer program code comprising instructions for segmenting a set of two-dimensional computed tomography (CT) slices corresponding to a lesion that, when executed by a processor, causes the processor to:

for each of at least a subset of the set of CT slices:
input the CT slice into a plurality of branches of a trained segmentation block, wherein each branch includes a convolutional neural network (CNN) with convolutional filters at a different scale, and each branch produces a plurality of levels of output;
generate feature maps for each of the levels of outputs, the feature maps generated based on a concatenation of same-level outputs across each of the branches;
and
generate a segmentation of the CT slice based on the feature maps of each of the levels of output;
aggregate the segmentations of each slice in the subset of CT slices to generate a three-dimensional segmentation of the lesion;
and
provide data representing the three-dimensional segmentation of the lesion for display of the three-dimensional segmentation.

5 more elements

A system for segmenting a set of two-dimensional computed tomography (CT) slices corresponding to a lesion comprising:

a processor;
and
a non-transitory computer-readable medium containing instructions that, when executed by the processor, cause the system to:
for each of at least a subset of the set of CT slices:
input the CT slice into a plurality of branches of a trained segmentation block, wherein each branch includes a convolutional neural network (CNN) with convolutional filters at a different scale, and each branch produces a plurality of levels of output;
generate feature maps for each of the levels of outputs, the feature maps generated based on a concatenation of same-level outputs across each of the branches;
and
generate a segmentation of the CT slice based on the feature maps of each of the levels of output;
aggregate the segmentations of each slice in the subset of CT slices to generate a three-dimensional segmentation of the lesion;
and
providing data representing the three-dimensional segmentation of the lesion for display of the three-dimensional segmentation.

8 more elements

Dependencies are inferred from the claim text.

Progressively-trained scale-invariant and boundary-aware deep neural network for the automatic 3D segmentation of lung lesions

Life of this patent

Progressively-trained scale-invariant and boundary-aware deep neural network for the automatic 3D segmentation of lung lesions

Life of this patent

Abstract

Claims20