Discrete Wavelet Transform: A breakthrough in segmentation of CT scans for Intracranial Hemorrhages

Abstract

Head injury is a major source for grimness and mortality worldwide and traumatic head wounds are a main source of neurological disability. Head wounds might go from a basic knock on the head to a skull crack and may cause cerebral harm and may even result in death. A traumatic brain injury happens when the skull is harmed, either due to an accident or an injury. This causes the blood to coagulate outside the brain matter within the skull or inside the brain matter itself which is recognized as Intracranial Haemorrhage. This is easily diagnosed using a CT scan of the brain. However, the CT scans may vary in complexity. To address the complexity of the brain CT images, this research paper suggests a method of intracranial Haemorrhage segmentation using discrete wavelet transform. The proposed method is based on a wavelet family used to help in extracting regions of Intracranial haemorrhages in the gray scale images and further applying morphological operations to denoise the image for better segmentation of the Haemorrhage. The proposed algorithm has achieved an Intersection Over Union Score of 78% and is tested on publicly available Kaggle’s Computed Tomography CT dataset to verify the segmented region.