Identification of Intracranial Hemorrhage using ResNeXt Model

Intracranial hemorrhage is a disease with a greater mortality rate. The only way to provide a definitive diagnosis of intra cranial hemorrhage is through neuroimaging. Deep learning tools can help identify and classify intracranial hemorrhages in a clinical setting and can also provide insights into previously unknown features of the disease. This paper summarizes state-of-the-art techniques for classification of intracranial hemorrhages using deep learning models and a novel model based on ResNext 101 is presented for interpretable diagnostic tool. The contributions of the paper include recommendations on pre-processing steps, model identification and training strategies and selection of datasets.