This paper presents an enhanced deep learning model for detecting pulmonary nodules in CT images, addressing the challenges of transfer learning due to discrepancies between source and target datasets. The proposed model integrates a pre-trained GoogLeNet Inception V3 network with a custom-designed feature fusion layer to improve feature extraction. The model was evaluated on the LUNA16 dataset, achieving an accuracy of 88.78% and sensitivity of 87.18%, which are improvements of 2.7% and 2.22% over the GoogLeNet Inception V3 algorithm. Further tests on different dataset proportions showed enhanced generalization capabilities. The model uses stochastic gradient descent for weight updates and incorporates data augmentation and fine-tuning to mitigate overfitting. The study also highlights the importance of feature fusion in improving performance and demonstrates the model's potential for clinical diagnosis. The results indicate that the enhanced model outperforms the original GoogLeNet Inception V3 model in terms of accuracy and convergence speed. The research underscores the effectiveness of deep learning in medical image recognition and suggests future directions for integrating computational features with semantic attributes for more comprehensive analysis of pulmonary nodules.This paper presents an enhanced deep learning model for detecting pulmonary nodules in CT images, addressing the challenges of transfer learning due to discrepancies between source and target datasets. The proposed model integrates a pre-trained GoogLeNet Inception V3 network with a custom-designed feature fusion layer to improve feature extraction. The model was evaluated on the LUNA16 dataset, achieving an accuracy of 88.78% and sensitivity of 87.18%, which are improvements of 2.7% and 2.22% over the GoogLeNet Inception V3 algorithm. Further tests on different dataset proportions showed enhanced generalization capabilities. The model uses stochastic gradient descent for weight updates and incorporates data augmentation and fine-tuning to mitigate overfitting. The study also highlights the importance of feature fusion in improving performance and demonstrates the model's potential for clinical diagnosis. The results indicate that the enhanced model outperforms the original GoogLeNet Inception V3 model in terms of accuracy and convergence speed. The research underscores the effectiveness of deep learning in medical image recognition and suggests future directions for integrating computational features with semantic attributes for more comprehensive analysis of pulmonary nodules.