An Optimized Model for Identification of Cerebral Palsy Using Deep Learning

Authors

  • Md Anjar Ahsan Faculty of Defence Science and Technology, National Defence University of Malaysia, Kuala Lumpur, Malaysia
  • Hassan bin Mohammed Cyber Security and Digital Industrial Revolution Centre, National Defence University of Malaysia, Kuala Lumpur, Malaysia
  • Abhilash Maroju Department of Information Technology, University of the Cumberlands USA, USA
  • Nurhafizah Moziyana Mohd Yusop Faculty of Defence Science and Technology, National Defence University of Malaysia, Kuala Lumpur, Malaysia
  • Wan Su Emi Yusnita Wan Yusof Department of Management, Faculty of Defence Studies and Management, National Defence University of Malaysia, Kuala Lumpur, Malaysia
  • Najjah Salwa Abd Razak Department of Language and Culture, Language Centre, National Defence University of Malaysia, Sungai Besi Camp, 57000, Kuala Lumpur, Malaysia
  • Mohd Arif Dar Center of Ionics, Department of Physic, University of Malaya, Kula Lampur 50603, Malaysia

Keywords:

Cerebral palsy classification, Functional MRI, Deep convolutional neural network, AlexNet architecture, Early diagnosis and rehabilitation

Abstract

Cerebral palsy (CP), a neurological disorder that affects children and can occasionally result in cognitive problems as well as deficits in motor function can be caused by prenatal, perinatal, or postnatal factors. Each subtype of cerebral palsy (CP), such as spastic and non-spastic cerebral palsy, has distinct symptoms based on the location of the brain lesion and how it affects muscle tone. Individualized therapy and rehabilitation programs are necessary to treat these differences effectively. Therefore, early-stage CP categorization is crucial to ensuring timely and targeted treatment efforts. The functional magnetic resonance imaging (fMRI) of the infant's brain is a helpful technique for CP imaging and early detection. This research uses a deep convolutional neural network (CNN) based on a modified AlexNet architecture to classify CP subtypes using newborn fMRI data. The modified AlexNet architecture gives an accuracy 79.5 % which is better than the results obtained through GoogleNet, AlexNET and LeNet models. This methodology aims to assist healthcare providers in developing more targeted recuperation programs, which will ultimately improve the lives of affected teenagers.

Downloads

Published

2025-07-04