Scholarlink

<p>Parkinson’s disease (PD) is a neurological condition marked by significant motor deficits resulting from damage to brain cells. It is essential to utilise MRI data to predict PD at an earlier stage, as current research often focuses on late disease stages, which limits the efficacy of early diagnosis approaches. This research represents a new effort to apply a novel method for early PD diagnosis, utilising a vast database of MRI scans. This methodology proposes a new deep architecture, the Concatenated Fusion-based Neural Network Strategy (CFNN). This integrates CNN models with advanced image analysis methods that incorporate both structural and functional features of the brain. It also involves preprocessing and normalizing MRI data to extract crucial features related to cortical thickness, volume, and connectivity patterns. A hybrid model that concatenates features, inspired by the fusion of multiple CNNs, is proposed to capture a broader spectrum of brain abnormalities associated with early PD. A novel optimization algorithm, Gradient-Free Adaptive Optimisation (GFAO), is used to fine-tune the model’s parameters to yield better performance after classifiers are trained with an adequate ensemble. The proposed CFNN-GFAO model, incorporating KNNs and DTs, achieved very high classification performance (up to 100% on the test set of this specific dataset), precision, recall, and F1-score, reflecting a 1.77% improvement over the base CFNN-GFAO model (98.23%). Compared to CFNN-GFAO with SVMs (99.57%), NBs (98.43%), and the Average Ensemble model (99.67%), the enhancements are 0.43%, 1.57%, and 0.33%, respectively.</p>

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

Gradient-free concatenated fusion neural network for early diagnosis of Parkinson’s disease using MRI data

  • 期刊论文

  • 发表日期：

    2026年6月28日

查看全文
  • Shaharyar Alam Ansari,
  • Shahnawaz Ahmad,
  • Mohammad Luqman,
  • Savir Ali,
  • Mohd. Farooq

摘要

Parkinson’s disease (PD) is a neurological condition marked by significant motor deficits resulting from damage to brain cells. It is essential to utilise MRI data to predict PD at an earlier stage, as current research often focuses on late disease stages, which limits the efficacy of early diagnosis approaches. This research represents a new effort to apply a novel method for early PD diagnosis, utilising a vast database of MRI scans. This methodology proposes a new deep architecture, the Concatenated Fusion-based Neural Network Strategy (CFNN). This integrates CNN models with advanced image analysis methods that incorporate both structural and functional features of the brain. It also involves preprocessing and normalizing MRI data to extract crucial features related to cortical thickness, volume, and connectivity patterns. A hybrid model that concatenates features, inspired by the fusion of multiple CNNs, is proposed to capture a broader spectrum of brain abnormalities associated with early PD. A novel optimization algorithm, Gradient-Free Adaptive Optimisation (GFAO), is used to fine-tune the model’s parameters to yield better performance after classifiers are trained with an adequate ensemble. The proposed CFNN-GFAO model, incorporating KNNs and DTs, achieved very high classification performance (up to 100% on the test set of this specific dataset), precision, recall, and F1-score, reflecting a 1.77% improvement over the base CFNN-GFAO model (98.23%). Compared to CFNN-GFAO with SVMs (99.57%), NBs (98.43%), and the Average Ensemble model (99.67%), the enhancements are 0.43%, 1.57%, and 0.33%, respectively.