Thyroid-associated ophthalmopathy(TAO)is a common chronic blinding orbital disease with poorly characterized early pathology.To solve the problem of misdiagnosis caused by subjective diagnostic bias of physicians in the TAO diagnosis process,a model Efficient-TAO based on the compound model scaling(CMS)method was proposed.Through optimizing the training process,the dynamic ReLU activation function,stochastic weight averaging function were improved to stimulate the perfor-mance.Direct training of ocular muscle CT image data was achieved by using image pre-processing algorithms.Experimental results show that the model achieves 98.34％classification accuracy and performs well in other algorithm evaluation metrics.The visualization results further validate the credibility of the Efficient-TAO algorithm for TAO classification.
维普
万方数据