Online continuous learning modular medical image labeling system capable of expanding tasks

• Main Authors: DING YUGUO, QIAO TIAN, ZHAI FANGWEN
• Format: Patent
• Language: Chinese; English
• Published: 16.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATIONTECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING ORPROCESSING OF MEDICAL OR HEALTHCARE DATA; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTEDFOR SPECIFIC APPLICATION FIELDS; PHYSICS

The invention provides an online continuous learning modular medical image labeling system capable of expanding tasks. The system comprises a data preprocessing module, a model pre-labeling/training module and an evaluation/interactive labeling module. Wherein the data preprocessing module is used for automatically acquiring a data set; and the model pre-labeling/training module is used for training the mounted neural network model in stages by using a data set in a database, and generating a pseudo label for an unlabeled data sample by using the trained neural network model to serve as a stage labeling result. And the evaluation/interaction labeling module is used for labeling samples or evaluating labeling quality in an interaction mode. Compared with the prior art, the method has the advantages that a modular structure is introduced, so that the system can share and utilize existing model weights, plug-and-play annotation among different tasks is realized, target tasks can be dynamically added, and annotat