Brain Haemorrhage Detection using LSTM, Convolution Neural Network and CT Scan Images

Abstract

A brain hemorrhage is an eruption of the  brain's arteries brought on by either excessive blood  pressure or blood coagulation, which may result in  fatalities or serious injuries. It is the kind of medical  emergency that requires a clinician to quickly identify the  site of internal bleeding before beginning therapy.  Convolutional Neural Network (CNN) and CNN + LSTM  hybrid models for deep learning are suggested in this study  for the categorization of brain hemorrhages. The 200 head  CT scan images dataset is utilized to increase the deep  learning models' precision and processing capability. Because big datasets are rarely immediately available in  essential situations, the main goal of this work is to apply  the abstraction capability of deep learning on a smaller amount of pictures A unique architecture called Brain  Haemorrhage Classification based on Neural Network was  made utilizing the CNN model together with picture  augmentation and dataset misbalancing approaches. The  performance of the recommended technique is assessed  using accuracy, precision, sensitivity, specificity, and F1- score. The experimental results are further evaluated  utilizing comparative analyses of the balanced and  unbalanced dataset using CNN and CNN + LSTM.  Unbalancing the dataset yields encouraging results,  outperforming CNN in accuracy. The results show the  effectiveness of the proposed method for quick  implementation in real-world circumstances and precise  prediction to preserve the patient's life in the interim.