US 12,482,103 B2
Tooth disease classification method based on federated learning (FL)
Xiaodong Li, Hangzhou (CN); Zhengsheng Yu, Hangzhou (CN); and Zhaozhe Gong, Hangzhou (CN)
Assigned to HANGZHOU DIANZI UNIVERSITY, Hangzhou (CN)
Filed by HANGZHOU DIANZI UNIVERSITY, Hangzhou (CN)
Filed on Feb. 24, 2023, as Appl. No. 18/173,875.
Claims priority of application No. 202210388126.5 (CN), filed on Apr. 13, 2022.
Prior Publication US 2023/0334668 A1, Oct. 19, 2023
Int. Cl. G06K 9/00 (2022.01); G06T 7/00 (2017.01); G06T 7/11 (2017.01); G06T 7/194 (2017.01); G16H 30/40 (2018.01)
CPC G06T 7/0014 (2013.01) [G06T 7/11 (2017.01); G06T 7/194 (2017.01); G16H 30/40 (2018.01); G06T 2207/10024 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/30036 (2013.01)] 10 Claims
OG exemplary drawing
 
1. A tooth disease classification method based on federated learning (FL), comprising the following steps:
acquiring oral information of a user;
transmitting an oral dental image to a local server for preprocessing;
performing, by the local server, model training according to a preprocessed oral dental image set and a model of a cloud master server;
transmitting trained model parameters to the cloud master server through digital encryption; and
performing, by each of local servers, fusion and classification according to digitally decrypted model parameters from the cloud master server;
wherein the second transmitting comprises:
outputting the model parameters in a form of the preprocessed oral dental image set, wherein output model parameters are mapped as binary data, an independent data matrix An is formed for each of teeth, output oral dental images are sequentially arranged to form a 6*6 matrix with respect to the An, and deficient bits are temporarily kept vacant;
supplementing an odd number of 0 in case of an odd number of the teeth, re-arranging the 6*6 matrix to complete the 6*6 matrix, and supplementing each of columns containing the temporarily vacant deficient bits by 0, wherein 0-supplementing positions are sequentially provided behind An having a maximum rank value in the column of An in the 6*6 matrix, and An under the 0-supplementing positions moves down sequentially; in case of two deficient bits in the column, the 0-supplementing positions are provided behind An having a second largest rank value in the column of An; and in case of a plurality of maximum rank values in the column of An, An on a last line of the column is taken;
supplementing 0 for re-arrangement in case of an even number of the teeth, wherein 0-supplementing positions are sequentially provided in front of An having a minimum rank value in each of columns of An in the 6*6 matrix, and An under the 0-supplementing positions moves down; in case of two deficient bits in the column, the 0-supplementing positions are provided in front of An having a second smallest rank value in the column of An; and in case of a plurality of minimum rank values in the column of An, An on a headmost line of the column is taken; and
recording the 0-supplementing positions to form a 0-supplementing matrix, inverting the 0-supplementing matrix to form an encryption matrix, and sending the model parameters to the cloud master server in the form of the encryption matrix.