Alcoholism Identification Based on an AlexNet Transfer Learning Model

(Aim) This paper proposed a novel alcoholism identification approach that can assist radiologists to make diagnosis. (Method) AlexNet was used as the basic transfer learning model. Global learning rate was small at 10-4, and the iteration epoch number as 10. The learning rate factor of replaced layers as 10 times larger than that of transferred layers. We tested five different replacement configurations of transfer learning. (Results) The experiment shows replacing the final fully connected layer achieved the best performance. Our method yielded a sensitivity of 97.44%± 1.15%, a specificity of 97.41%± 1.51%, a precision of 97.34%± 1.49%, an accuracy of 97.42%± 0.95%, and an F1 score of 97.37%± 0.97% on the test set. (Conclusion) This method can assist radiologists in their routine alcoholism screening of brain magnetic resonance images.