Computer Aided Diagnosis Model for Haemorrhage Detection using Improved Equilibrium Optimization with Deep Learning on CT Images

Abstract

Intracranial haemorrhage (ICH) recognition is a crucial challenge in neurology and radiology, as on time detection of haemorrhages within brain help in fast involvement and treatment. Numerous image modalities, with magnetic resonance imaging (MRI) and computed tomography (CT) are widely utilized to perceive and categorize ICH. Traditional models for ICH recognition are trusted on physical assessment of CT images by expert radiologists. But, with developments in machine learning (ML), deep learning (DL) and computer aided methods are mainly established to aid radiologists in identifying and detecting ICH professionally. DL methods like convolutional neural networks (CNN) have revealed excellent outcomes in ICH recognition on CT images. This article presents Improved Equilibrium Optimization with Deep Learning for ICH detection (IEODL-ICHD) technique on CT images. The IEODL-ICHD technique starts with the UNet segmentation approach which proficiently outlines the regions of interest for ICH detection. Besides, the IEODL-ICHD technique utilizes the ResNet18 feature extractor to extract high-level features to accurately identify the haemorrhage. Furthermore, the IEO model is exploited for hyperparameter tuning, safeguarding the model's flexibility to the refined features of the ICH. Lastly, the extreme gradient boosting (XGBoost) model is applied for accurate recognition and classification of the ICH. The simulation results of the IEODL-ICHD technique are systematically authenticated on CT images, representing greater performance in ICH classification.