Intracranial Hemorrhage Detection and Classification of its Subtypes using CNN, ResNet, Random Forest and DenseNet201

Abstract

Intracranial haemorrhage is one of the most life-threatening conditions in which a person bleeds from or within the brain tissue and the skull. There are several types of intracranial haemorrhage like epidural haemorrhage, subdural haemorrhage, interventricular haemorrhage, intracerebral haemorrhage and lastly subarachnoid haemorrhage. This problem should be detected and treated as soon as possible otherwise it can cause death. A system is required in order to accurately identify patients and expedite the healing process. This research article proposes OT-DLRF (Otsu thresholding and deep learning with random forest classification) model for diagnosing and classifying the Intracranial haemorrhage. The performance of the model is dependent upon the combination model of CNN and DenseNet 201 and for the classification of the final output the Random Forest (RF) classifier is used. Model shows an accuracy of 99.6 percent for subdural, epidural, intra-parenchymal haemorrhage.