18-07-2023 дата публикации
Номер: CN116452617A
Принадлежит:
The invention discloses a whole heart segmentation method based on cross-modal domain transfer learning. The method comprises the following steps of S1, performing data preprocessing on an initial whole heart MRI image and an initial CT image; s2, inputting the preprocessed image into a modal migration network MTN for data training, and generating a new CT image; s3, inputting the new CT image generated by the modal migration network MTN and the CT heart image with the all-heart label into a U-shaped multi-attention network MAUNet, and outputting an intermediate heart image segmentation result; and S4, inputting the intermediate heart image segmentation result into a spatial configuration network SCN for data training, and outputting a final heart image segmentation result. According to the method, the heart is segmented by using the multi-modal data, the multi-modal data has multi-level information due to different imaging mechanisms, and when the same tissue is segmented by using the ...
Подробнее