Abstract: In order to solve the problem of uncertain quantification in the classification of pulmonary nodules, starting from practical clinical needs and technology, realize the value application of its feature information. An uncertainty and multi-stream dual-branch feature fusion collaborative optimization pulmonary nodule classification method(UC-MDFUN) is proposed. The three-dimensional image and content information of uncertain nodules were guided to realize cascade adaptive weighted fusion of radiomics and multi-stream deep learning advanced features, improve information diversity and richness, improve model confidence and wide pulmonary nodule classification application ability, malignant recognition ability, prediction and network generalization ability. The experimental results show that the sensitivity and AUC of the algorithm are 96.3% and 96.1%, and it has good classification performance and ability to identify malignant nodules.