Brain MR Image Classification Using Superpixel-Based Deep Transfer Learning.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Deep Learning for Diagnostics

Background:

• Brain Magnetic Resonance (MR) imaging is crucial for diagnosing neurological conditions.
• Classifying MR images as normal or abnormal is essential for patient care.
• Existing methods may face challenges with small-scale medical datasets.

Purpose of the Study:

• To develop an efficient deep learning architecture for brain MR image classification.
• To leverage superpixel segmentation and transfer learning for improved diagnostic accuracy.
• To evaluate the proposed model on multiple brain MR image datasets.

Main Methods:

• An ensemble transfer learning-inspired deep architecture was designed.
• Simple Linear Iterative Clustering (SLIC) superpixel algorithm was used for image segmentation.
• Superpixel images were fed into a Convolutional Neural Network (CNN) for classification.

Main Results:

• The model achieved high accuracy across three datasets: 88.15% (NITR-DHH), 98.15% (DS-160), and 98.33% (DS-75).
• Effective performance was demonstrated even with small-scale medical image datasets.
• The proposed method proved to be promising and efficient for clinical applications.

Conclusions:

• The proposed SLIC-based superpixel and CNN transfer learning model is effective for brain MR image classification.
• This approach offers a viable solution for diagnosing brain pathologies using medical imaging.
• The method shows significant potential for enhancing clinical diagnostic capabilities.