Scholarlink

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that affects millions of people worldwide. This study presents the development of a neural network model for detecting PD using an improved architecture of the AlexNet convolutional neural network (CNN). Magnetic resonance imaging (MRI) data was from the Parkinson’s Progression Markers Initiative (PPMI) database. From an initial set of 27,270 MRI images, 4000 images were selected for model training and analysis, which included data from 103 patients with PE and 24 healthy controls. To allow efficient processing, DICOM files were pre-processed and then converted to PNG format. This step ensured support for advanced image manipulation techniques and seamless integration into deep learning workflows. The training and validation process, conducted with the TensorFlow and Keras libraries, resulted in a model that can distinguish PE patients from healthy individuals with an accuracy of 92%. The results of this research highlight the potential of deep learning techniques to aid in early diagnosis and PE intervention, offering valuable insights into advancing the management of neurodegenerative diseases.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced AlexNet CNN for Accurate Detection of Parkinson’s Disease Using MRI Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ariel Miranda-Carrillo,
  • Camila Usama-Huertas,
  • Kevin Samaniego-Molina,
  • Ricardo Moreira Intriago,
  • Fernando Villalba-Meneses,
  • Paulo Navas,
  • Andrés Tirado-Espín,
  • Paulina Vizcaíno-Imacaña,
  • Carolina Cadena-Morejón,
  • Diego Almeida-Galárraga

摘要

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that affects millions of people worldwide. This study presents the development of a neural network model for detecting PD using an improved architecture of the AlexNet convolutional neural network (CNN). Magnetic resonance imaging (MRI) data was from the Parkinson’s Progression Markers Initiative (PPMI) database. From an initial set of 27,270 MRI images, 4000 images were selected for model training and analysis, which included data from 103 patients with PE and 24 healthy controls. To allow efficient processing, DICOM files were pre-processed and then converted to PNG format. This step ensured support for advanced image manipulation techniques and seamless integration into deep learning workflows. The training and validation process, conducted with the TensorFlow and Keras libraries, resulted in a model that can distinguish PE patients from healthy individuals with an accuracy of 92%. The results of this research highlight the potential of deep learning techniques to aid in early diagnosis and PE intervention, offering valuable insights into advancing the management of neurodegenerative diseases.