公开(公告)号：US20210401392A1

公开(公告)日：2021-12-30

申请号：US17473495

申请日：2021-09-13

Applicant: GENENTECH, INC.

Inventor： Nils Gustav Thomas BENGTSSON ,  Richard Alan Duray CARANO ,  Alexander James DE CRESPIGNY ,  Jill O. FREDRICKSON ,  Mohamed Skander JEMAA

IPC: A61B6/00 ,  G06T7/174 ,  G06T7/11 ,  G16H30/20 ,  G16H30/40 ,  G16H50/20 ,  G16H50/30 ,  G16H10/20 ,  G16H50/50 ,  G06N3/02

Abstract: The present disclosure relates to techniques for segmenting tumors with positron emission tomography (PET) using deep convolutional neural networks for image and lesion metabolism analysis. Particularly, aspects of the present disclosure are directed to obtaining a PET scans and computerized tomography (CT) or magnetic resonance imaging (MRI) scans for a subject, preprocessing the PET scans and the CT or MRI scans to generate standardized images, generating two-dimensional segmentation masks, using two-dimensional segmentation models implemented as part of a convolutional neural network architecture that takes as input the standardized images, generating three-dimensional segmentation masks, using three-dimensional segmentation models implemented as part of the convolutional neural network architecture that takes as input patches of image data associated with segments from the two-dimensional segmentation mask, and generating a final imaged mask by combining information from the two-dimensional segmentation masks and the three-dimensional segmentation masks.

公开(公告)号：US20240303822A1

公开(公告)日：2024-09-12

申请号：US18669407

申请日：2024-05-20

Applicant: Genentech, Inc.

Inventor： Mohamed Skander JEMAA ,  Yury Anatolievich PETROV ,  Xiaoyong WANG ,  Nils Gustav Thomas BENGTSSON ,  Richard Alan Duray CARANO

IPC: G06T7/11 ,  G06T7/33

CPC classification number: G06T7/11 ,  G06T7/337 ,  G06T2207/10081 ,  G06T2207/20016 ,  G06T2207/20084 ,  G06T2207/20221 ,  G06T2207/30096

Abstract: In one embodiment, a method includes accessing a first scan image from a set of computed tomography (CT) scan images with each CT scan image being at a first resolution, generating a first downscaled image of the first scan image by resampling the first scan image to a second resolution that is lower than the first resolution, determining coarse segmentations corresponding to organs portrayed in the first scan image by first machine-learning models based on the first downscaled image, extracting segments of the first scan image based on the coarse segmentations with each extracted segment being at the first resolution, determining fine segmentations corresponding to the respective organs portrayed in the extracted segments by second machine-learning models based on the extracted segments, and generating a segmented image of the first scan image based on the fine segmentations, wherein the segmented image comprises confirmed segmentations corresponding to the organs.

公开(公告)号：US20250000473A1

公开(公告)日：2025-01-02

申请号：US18883384

申请日：2024-09-12

Applicant: GENENTECH, INC.

Inventor： Nils Gustav Thomas BENGTSSON ,  Richard Alan Duray CARANO ,  Alexander James Stephen CHAMPION DE CRESPIGNY ,  Jill Osborn FREDRICKSON ,  Mohamed Skander JEMAA

IPC: A61B6/00 ,  G06N3/02 ,  G06T7/11 ,  G06T7/174 ,  G16H10/20 ,  G16H30/20 ,  G16H30/40 ,  G16H50/20 ,  G16H50/30 ,  G16H50/50

Abstract: The present disclosure relates to techniques for segmenting tumors with positron emission tomography (PET) using deep convolutional neural networks for image and lesion metabolism analysis. Particularly, aspects of the present disclosure are directed to obtaining a PET scans and computerized tomography (CT) or magnetic resonance imaging (MRI) scans for a subject, preprocessing the PET scans and the CT or MRI scans to generate standardized images, generating two-dimensional segmentation masks, using two-dimensional segmentation models implemented as part of a convolutional neural network architecture that takes as input the standardized images, generating three-dimensional segmentation masks, using three-dimensional segmentation models implemented as part of the convolutional neural network architecture that takes as input patches of image data associated with segments from the two-dimensional segmentation mask, and generating a final imaged mask by combining information from the two-dimensional segmentation masks and the three-dimensional segmentation masks.

公开(公告)号：US20230025980A1

公开(公告)日：2023-01-26

申请号：US17782497

申请日：2020-12-04

Applicant: Genentech, Inc.

Inventor： Michael Gregg KAWCZYNSKI ,  Jeffrey R. WILLIS ,  Nils Gustav Thomas BENGTSSON ,  Jian DAI ,  Simon Shang GAO

IPC: G06T7/00 ,  G06T3/00 ,  G06T7/187

Abstract: Systems and methods relate to processing optical tomography coherence (OCT) images to predict characteristics of a treatment to be administered to effectively treat age-related macular degeneration. The processing can include pre-processing the image by flattening and/or cropping the image and processing the pre-processed image using a neural network. The neural network can include a deep convolutional neural network. An output of the neural network can indicate a predicted frequency and/or interval at which a treatment (e.g., anti-vascular endothelial growth factor therapy) is to be administered so as to prevent leakage of vasculature in the eye.

公开(公告)号：US20220319008A1

公开(公告)日：2022-10-06

申请号：US17842542

申请日：2022-06-16

Applicant: Genentech, Inc.

Inventor： Nils Gustav Thomas BENGTSSON ,  Richard Alan Duray CARANO ,  Alexander James Stephen CHAMPION DE CRESPIGNY ,  Jill Osborn FREDRICKSON ,  Mohamed Skander JEMAA

IPC: G06T7/00 ,  G06T7/12 ,  G06T7/70 ,  G06T7/62 ,  A61B5/00

Abstract: Medical image(s) are input into a detection network to generate mask(s) identifying a set of regions within the medical image(s), where the detection network predicts that each region identified in the mask(s) includes a depiction of a tumor of one or more tumors within the subject. For each region, the region of the medical image(s) is processed using a tumor segmentation network to generate one or more tumor segmentation boundaries for the tumor present within the subject. For each tumor and by using a plurality of organ-specific segmentation networks, an organ is determined within which at least part of the tumor is located. An output is generated based on the one or more tumor segmentation boundaries and locations of the organs within which at least part of the one or more tumors are located.