Optimization and Application of Convolutional Neural Networks in Medical Imaging Data Analysis
DOI:
https://doi.org/10.53469/jtpms.2025.05(4).08Keywords:
Convolutional Neural Network, Medical imaging analysis, Deep learning, Pulmonary nodule detectionAbstract
The purpose of this study is to improve the performance of convolutional neural networks in medical imaging data analysis by optimizing their architecture and loss function. Method: Two publicly available datasets, LIDC-IDRI and BraTS, were selected for experiments on lung nodule detection and brain tumor segmentation tasks. Introducing deep network structures such as ResNet and combining loss functions such as Dice Loss and Focal Loss to optimize the model, while utilizing cross validation and early stopping strategies to enhance the model's generalization ability. Result: The accuracy of the optimized model on the LIDC-IDRI dataset has increased to 95.2%, while its sensitivity and Dice coefficient have reached 93.5% and 0.894, respectively; On the BraTS dataset, we successfully increased the accuracy to 94.1%, while also achieving a Dice coefficient of 0.885. Conclusion: Optimizing the CNN model significantly improves the accuracy of lesion area detection and segmentation, demonstrating its potential for application in medical imaging data analysis.
References
Junhui Li Multi Objective Evolutionary Architecture Search of Convolutional Neural Networks for Medical Image Segmentation [D]. South China University of Technology, 2021.
Qi Zhang Research on Medical Image Recognition Algorithm Based on Convolutional Neural Network [D]. Hebei University of Economics and Business, 2021.
