Multi-Class Classification of Intracranial Hemorrhages in a 3-Channel CT image by using a Transfer Learning based DenseNet121 model

Immediate detection of the kind of intracranial hemorrhage (ICH) and subsequent treatment is required for individuals with brain hemorrhages to have a better chance of survival. Deep learning models have been demonstrated to be extremely competent in supporting doctors in the classification of cerebral hemorrhages. This paper evaluates the DenseNet121 model using the transfer learning concept. The database consists of a total of 26383 CT images with different types of ICH including normal scans. The performance results on the validation data were an accuracy of 94.8%, a precision of 94.9%, a recall of 81.4%, an F1 score of 87.2%, and a ROC under AUC of 99.1% respectively. The results ensure that the proposed model has reached a higher accuracy when compared to the other models.