CEREBRAL MICRO-BLEEDING SEGMENTATION BASED ON CONVOLUTIONAL NEURAL NETWORKS (CNN)

  • T. Suleiman Department of Computer Science and IT, Federal University Dutsinma
  • U. Iliyasu Department of Computer Science and IT, Federal University Dutsinma
  • A. J. Abdul Department of Computer Science, Ahmadu Bello University, Zaria

Abstract

Manual detection of Cerebral Microbleeds (CMB) by radiologists is time consuming and prone to errors. Hence, researchers have developed various Machine Learning techniques to automate this task. The highest performing model on the CMB dataset using Convolutional Neural Network (CNN) achieved a prediction accuracy of 97.78%. The importance of accuracy especially in diagnosis cannot be over emphasized as the patient has so much to lose when there is a wrong diagnosis. These models have been localized by the researchers as they were not deployed to enable radiologists’ means to carry out CMB detection remotely. This project was aimed at achieving an improved performance in CMB detection by changing the structure and parameter settings of the CNN model. The final model developed from this project achieved a record accuracy of 98.8% in predicting CMBs.

Published
2021-12-22
How to Cite
SULEIMAN, T.; ILIYASU, U.; ABDUL, A. J.. CEREBRAL MICRO-BLEEDING SEGMENTATION BASED ON CONVOLUTIONAL NEURAL NETWORKS (CNN). SAU Science-Tech Journal, [S.l.], v. 6, n. 1, p. 130-134, dec. 2021. ISSN 2659-1529. Available at: <https://journals.sau.edu.ng/index.php/sjbas/article/view/505>. Date accessed: 28 june 2022.
Section
Articles