Prediction of uterine fibroid outcomes based on DenseNet and multiple MRI sequences

High-intensity focused ultrasound (HIFU) is a surgical method commonly used to treat uterine fibroids, but not all patients are suitable for this treatment. Therefore, effective postoperative outcome prediction can improve surgical success rate. In this study, a data-driven prediction model for HIFU treatment outcome was developed based on patients' MRI sequences and DenseNet network. We used N4 bias correction, SimpleITK intensity normalization, image registration and other preprocessing methods to improve image quality. We also introduced data augmentation and transfer learning methods to solve the problem of small dataset and integrated multimodal information through feature fusion method. In conclusion, the fusion model's Accuracy was 0.769, AUC value was 0.806, which was better than the prediction accuracy of single sequence, and the experimental results showed that the model had good predictive performance.