公开(公告)号：US12040080B2

公开(公告)日：2024-07-16

申请号：US17620445

申请日：2020-09-11

Applicant: Google LLC

Inventor： Robert Carter Dunn ,  Ayush Jain ,  Peggy Yen Phuong Bui ,  Clara Eng ,  David Henry Way ,  Kang Li ,  Vishakha Gupta ,  Jessica Gallegos ,  Dennis Ai ,  Yun Liu ,  David Coz ,  Yuan Liu

IPC: G16H30/40 ,  G16H10/60 ,  G16H50/20

CPC classification number: G16H30/40 ,  G16H10/60 ,  G16H50/20

Abstract: The present disclosure is directed to a deep learning system for differential diagnoses of skin diseases. In particular, the system performs a method that can include obtaining a plurality of images that respectively depict a portion of a patient's skin. The method can include determining, using a machine-learned skin condition classification model, a plurality of embeddings respectively for the plurality of images. The method can include combining the plurality of embeddings to obtain a unified representation associated with the portion of the patient's skin. The method can include determining, using the machine-learned skin condition classification model, a skin condition classification for the portion of the patients skin, the skin condition classification produced by the machine-learned skin condition classification model by processing the unified representation, wherein the skin condition classification identifies one or more skin conditions selected from a plurality of potential skin conditions.

公开(公告)号：US20240119586A1

公开(公告)日：2024-04-11

申请号：US17768419

申请日：2020-10-13

Applicant: Google LLC

Inventor： Vivek Natarajan ,  Yuan Liu ,  David Coz ,  Amirata Ghorbani

IPC: G06T7/00 ,  G06T11/00 ,  G16H10/60 ,  G16H50/70

CPC classification number: G06T7/0012 ,  G06T11/001 ,  G16H10/60 ,  G16H50/70 ,  G06T2207/20081 ,  G06T2207/20084 ,  G06T2207/20104 ,  G06T2207/30088 ,  G06T2207/30096

Abstract: We disclose the generation and training of Generative Adversarial Networks (GAN) to synthesize clinical images with skin conditions. Synthetic images for a pre-specified skin condition are generated, while being able to vary its size, location and the underlying skin color. We demonstrate that the generated images are of high fidelity using objective GAN evaluation metrics. The synthetic images are not only visually similar to real images, but also embody the respective skin conditions. Additionally, synthetic skin images can be used as a data augmentation technique for training a skin condition classifier, and improve the ability of the classifier to detect rare but malignant conditions.

公开(公告)号：US20220359062A1

公开(公告)日：2022-11-10

申请号：US17620445

申请日：2020-09-11

Applicant: Google LLC

Inventor： Robert Carter Dunn ,  Ayush Jain ,  Peggy Yen Phuong Bui ,  Clara Eng ,  David Henry Way ,  Kang Li ,  Vishakha Gupta ,  Jessica Gallegos ,  Dennis Ai ,  Yun Liu ,  David Coz ,  Yuan Liu

IPC: G16H30/40 ,  G16H50/20 ,  G16H10/60

Abstract: The present disclosure is directed to a deep learning system for differential diagnoses of skin diseases. In particular, the system performs a method that can include obtaining a plurality of images that respectively depict a portion of a patient's skin. The method can include determining, using a machine-learned skin condition classification model, a plurality of embeddings respectively for the plurality of images. The method can include combining the plurality of embeddings to obtain a unified representation associated with the portion of the patient's skin. The method can include determining, using the machine-learned skin condition classification model, a skin condition classification for the portion of the patients skin, the skin condition classification produced by the machine-learned skin condition classification model by processing the unified representation, wherein the skin condition classification identifies one or more skin conditions selected from a plurality of potential skin conditions.