US 11,710,231 B2
Method and apparatus for mammographic multi-view mass identification
Zhicheng Yang, Palo Alto, CA (US); Zhenjie Cao, Palo Alto, CA (US); Yanbo Zhang, Palo Alto, CA (US); Peng Chang, Palo Alto, CA (US); Mei Han, Palo Alto, CA (US); and Jing Xiao, Palo Alto, CA (US)
Assigned to PING AN TECHNOLOGY (SHENZHEN) CO., LTD., Shenzhen (CN)
Filed by Ping An Technology (Shenzhen) Co., Ltd., Shenzhen (CN)
Filed on Feb. 2, 2021, as Appl. No. 17/165,087.
Claims priority of provisional application 63/072,379, filed on Aug. 31, 2020.
Prior Publication US 2022/0067927 A1, Mar. 3, 2022
Int. Cl. G06T 7/00 (2017.01); G06T 7/11 (2017.01); G06T 7/30 (2017.01); G06V 10/25 (2022.01); G06V 10/75 (2022.01)
CPC G06T 7/0012 (2013.01) [G06T 7/11 (2017.01); G06T 7/30 (2017.01); G06V 10/25 (2022.01); G06V 10/751 (2022.01); G06T 2207/20076 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30068 (2013.01); G06V 2201/03 (2022.01); G06V 2201/07 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method, applied to an apparatus for mammographic multi-view mass identification, comprising:
receiving a main image, a first auxiliary image, and a second auxiliary image, wherein the main image and the first auxiliary image are images of a breast of a person, and the second auxiliary image is an image of another breast of the person;
detecting a nipple location based on the main image and the first auxiliary image;
generating a first probability map of the main image based on the main image, the first auxiliary image, and the nipple location by an ipsilateral analyzer that is built on a Faster-RCNN detection architecture and configured for:
enabling two input branches to share same weights and extract features from the main image and the first auxiliary image in a same way, and
modelling relationships between two region of interests (RoIs) in a single image based on similarities in brightness, shapes, area sizes, and relative positions of the RoIs,
wherein a relative position of a RoI includes a distance from the RoI to the nipple location, and in a training process, focal loss and distance-intersection-over-union loss (DIoU) is used to improve performance of the ipsilateral analyzer;
generating a second probability map of the main image based on the main image, the second auxiliary image, and the nipple location; and
generating and outputting a fused probability map based on the first probability map and the second probability map.