Fetal Congenital Heart Disease Diagnosis Based on CBAM-Enhanced ResNet-50

Fetal congenital heart disease refers to structural abnormalities in the heart or major blood vessels that occur during the fetal development stage. In order to detect and treat fetal lesions as early as possible, image classification is used to distinguish between normal and abnormal data sets in ultrasound imaging, which can assist doctors to make more precise diagnoses. Despite the challenges in classifying ultrasound images of fetal congenital heart disease, this study focuses on exploring the use of the CBAM-enhanced ResNet-50 model in diagnosing fetal congenital heart disease. To preserve the original structure of ResNet-50 and utilize its pre-trained parameters, we added the CBAM module after the last convolutional layer of ResNet-50, employing attention mechanisms to enhance the model’s focus on key features in fetal ultrasound images. Experimental results show that compared with the ResNet-50, introducing CBAM can significantly improve performance. The image classification accuracy improves by 0.84%, 0.09% and 1.96% when using spatial attention, channel attention and joint attention respectively in medical image classification.