Medical image report multi-label classification method based on deep learning

The invention discloses a medical image report multi-label classification method based on deep learning. The method comprises the steps of obtaining text data of a medical image report and label data of an abnormal condition in the medical image report; word vectors and label vectors are obtained according to the text data and the label data; inputting the word vector and the label vector into a multi-label classification model to obtain multi-label classification task loss; inputting the word vector, the text data and the label vector into an auxiliary task model to obtain auxiliary task loss; performing back propagation optimization on the multi-label classification model according to the auxiliary task loss and the multi-label classification task loss to obtain a training multi-label classification model; obtaining to-be-detected text data of a to-be-detected medical image report; and inputting the to-be-detected text data into the training multi-label classification model, and determining an abnormal prob