Related Experiment Video
Updated: Nov 18, 2025
15:53
Utilizing 3D Printing Technology to Merge MRI with Histology: A Protocol for Brain Sectioning
Published on: December 6, 2016
15.3K
Wilson disease tissue classification and characterization using seven artificial intelligence models embedded with 3D
Mohit Agarwal1, Luca Saba2, Suneet K Gupta1
1CSE Department, Bennett University, Greater Noida, UP, India.
Medical & Biological Engineering & Computing
|February 6, 2021
Summary
Artificial intelligence (AI) aids early diagnosis of Wilson's disease (WD) by analyzing brain MRI scans. An improved deep convolution neural network (iDCNN) shows superior accuracy in identifying WD white matter hyperintensities.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Wilson's disease (WD) involves toxic copper accumulation, primarily in the brain and liver, potentially leading to severe disability or death if untreated.
- Diagnosis of WD is challenging due to subtle white matter hyperintensity (WMH) changes on MRI and limitations in AI training data.
- Current diagnostic methods may lack the sensitivity for early detection of subtle WMH in WD patients.
Purpose of the Study:
- To design and validate advanced AI-based computer-aided design (CADx) systems for the classification and characterization of WD.
- To develop and compare a conventional deep convolution neural network (cDCNN) with an improved version (iDCNN) for enhanced WD diagnosis.
- To evaluate the performance of AI systems using metrics such as accuracy, AUC, and diagnostic odds ratio (DOR).
Main Methods:
- Developed seven AI-based CADx systems, including 3D optimized classification and characterization models.
- Proposed a modified iDCNN with a differentiable rectified linear unit (ReLU) activation function for improved performance.
- Utilized CNN-based feature map strength and Bispectrum analysis of pixels for WD characterization, optimizing CNN layers and augmentation folds.
Main Results:
- The iDCNN achieved superior performance with an accuracy of 98.28 ± 1.55% and an AUC of 0.99 (p < 0.0001).
- iDCNN demonstrated a fourfold higher DOR compared to cDCNN, indicating significantly improved diagnostic power.
- iDCNN outperformed transfer learning (Inception V3) by 11.92% and conventional machine learning models (k-NN, decision tree, SVM, random forest) by up to 55.13%.
Conclusions:
- AI-based CADx systems, particularly the iDCNN, show significant potential for the early and accurate diagnosis of Wilson's disease.
- The developed iDCNN model offers enhanced diagnostic capabilities for identifying subtle WMH associated with WD.
- These advanced AI tools could improve patient outcomes by facilitating timely intervention and treatment for Wilson's disease.
