US 12,094,188 B2
Methods and systems for training learning network for medical image analysis
Junhuan Li, Shenzhen (CN); Ruoping Li, Shenzhen (CN); Ling Hou, Shenzhen (CN); Pengfei Zhao, Shenzhen (CN); Yuwei Li, Bellevue, WA (US); Kunlin Cao, Kenmore, WA (US); and Qi Song, Seattle, WA (US)
Assigned to SHENZHEN KEYA MEDICAL TECHNOLOGY CORPORATION, Shenzhen (CN)
Filed by Shenzhen Keya Medical Technology Corporation, Shenzhen (CN)
Filed on Dec. 29, 2021, as Appl. No. 17/565,274.
Claims priority of application No. 202110525922.4 (CN), filed on May 13, 2021.
Prior Publication US 2022/0366679 A1, Nov. 17, 2022
Int. Cl. G06V 10/774 (2022.01); G06T 7/00 (2017.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/7747 (2022.01) [G06T 7/0012 (2013.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06T 2207/10081 (2013.01); G06T 2207/30048 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A training method for training a learning network for medical image analysis, comprising:
when a pre-trained learning network trained using an original training data set has an evaluation defect, performing, by a processor, a data augmentation on the original training data set for the evaluation defect to obtain a data augmented training data set; and
performing, by the processor, a refined training on the pre-trained learning network using the data augmented training data set,
wherein the refined training is performed in N stages with N being a positive integer, and when N is greater than 1, an Nth stage of the refined training comprises:
performing a data augmentation on a (N−1)th training data set for a Nth evaluation defect to obtain a Nth training data set; and
performing the Nth stage of the refined training on a (N−1)th learning network using the Nth training data set to obtain an Nth learning network.