US 11,941,813 B2
Systems and methods for performing segmentation based on tensor inputs
Bing Song, La Canada, CA (US); Nicholas James Witchey, Laguna Hills, CA (US); Albert Wu, Calabasas, CA (US); Krsto Sbutega, Redondo Beach, CA (US); and Patrick Soon-Shiong, Los Angeles, CA (US)
Assigned to NantCell, Inc., Culver City, CA (US)
Filed by NantCell, Inc., Culver City, CA (US)
Filed on Aug. 18, 2020, as Appl. No. 16/996,068.
Claims priority of provisional application 62/890,767, filed on Aug. 23, 2019.
Prior Publication US 2021/0056363 A1, Feb. 25, 2021
Int. Cl. G06N 3/0455 (2023.01); G06F 18/21 (2023.01); G06F 18/211 (2023.01); G06F 18/241 (2023.01); G06N 3/084 (2023.01); G06N 5/046 (2023.01); G06T 7/10 (2017.01); G06T 7/11 (2017.01); G06T 7/73 (2017.01); G06V 20/69 (2022.01)
CPC G06T 7/11 (2017.01) [G06F 18/211 (2023.01); G06F 18/2163 (2023.01); G06F 18/241 (2023.01); G06N 3/0455 (2023.01); G06N 3/084 (2013.01); G06N 5/046 (2013.01); G06T 7/10 (2017.01); G06T 7/74 (2017.01); G06V 20/695 (2022.01); G06V 20/698 (2022.01); G06T 2207/20081 (2013.01); G06T 2210/22 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A system for performing classification of data based on tensor inputs, the system comprising:
memory storing computer-executable instructions defining a learning network, where the learning network includes a plurality of sequential encoder down-sampling blocks, and the learning network is a segmentation model including deep learning parameters adjusted by comparing segmentation classifications produced by processing cropped training patches using the segmentation model to corresponding specified segmentation classification masks; and
a processor in communication with the memory, the processor configured to execute the computer-executable instructions to:
stack multiple biological cell images to form a multi-dimensional input tensor, the multiple biological cell images including biological cell images including different spectral bands of light, the multi-dimensional input tensor including at least a first dimension, a second dimension and a plurality of channels, wherein each of the different spectral bands of light is a channel C of the multi-dimensional input tensor, where parameters W and H of the multi-dimensional input tensor represent a width and a height of the biological cell images;
process the received multi-dimensional input tensor by passing the received multi-dimensional input tensor through the plurality of sequential encoder down-sampling blocks, the processing including a combination of cell segmentation and classification with multiple cell type annotation labels; and
generate an output tensor in response to processing the received multi-dimensional input tensor via the plurality of sequential encoder down-sampling blocks of the learning network, the output tensor including at least one segmentation classification.