OVARIAN CYSTADENOMA CT IMAGE LESION SEGMENTATION USING CONVOLUTIONAL NEURAL NETWORK

Ovarian cystadenoma is a disease that occurs in the ovary. In order to achieve automatic lesion segmentation of ovarian CT images, this paper proposes a CT image lesion segmentation method for ovarian cystadenoma based on an improved U-net model, using VGG16 as an encoder to further simplifying the structure of the U-net model, and combining the CT image features for data enhancement. This paper constructed an ovarian CT image dataset based on clinical diagnostic data for model training and evaluation. The model achieved an IoU of 88.85% and an AUC of 99.72% on the testing set, demonstrating the feasibility and accuracy of the improved U-net model for segmentation of ovarian cystadenoma lesions. Compared with the original U-net model, the proposed improvement can reduce the model size without losing segmentation accuracy and is more suitable for assisting clinical diagnosis.