Classification of Acute Bilirubin Encephalopathy Based on Individual and Fused Attention Network

Hyperbilirubinemia is a common disease caused by abnormal metabolism of bilirubin. As a serious complication of hyperbilirubinemia, acute bilirubin encephalopathy (ABE) in newborns, is caused by abnormal bilirubin metabolism leading to high levels of bilirubin in the blood. In severe cases, it can threaten the life of patients. Recently, medical image-assisted diagnosis technology based on convolutional neural networks has been widely used to assist radiologists in analyzing and diagnosing case samples of brain diseases. However, the existed methods are not good at representing the crucial information inside each modal and between modals. In this paper, a novel attention network based on individual attention module (IAM) and fused attention module (FAM) is proposed for acute bilirubin encephalopathy classification. The IAM is used to gather internal correlations and key information within each modality while the FAM is exploited to fuse the guided features from different modalities to enhance their connectivity. Experimental results indicate that the proposed network performs better in the medical image classification and diagnosis of ABE.