Related Experiment Video
Updated: Aug 9, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.0K
Feature Extraction Using a Residual Deep Convolutional Neural Network (ResNet-152) and Optimized Feature Dimension
Suganya Athisayamani1, Robert Singh Antonyswamy2, Velliangiri Sarveshwaran2
1School of Computing, Sastra Deemed to be University, Thanjavur 613401, India.
Diagnostics (Basel, Switzerland)
|February 25, 2023
Summary
This study introduces a novel, non-invasive method for classifying brain tumors using MRI scans. The proposed computational approach achieves high accuracy, improving diagnostic efficiency for brain cancer detection.
Area of Science:
- Medical imaging
- Computational oncology
- Artificial intelligence in healthcare
Background:
- Brain tumors are a leading cause of global mortality, with traditional biopsy methods facing limitations like low sensitivity and procedural risks.
- Accurate and timely classification of brain tumors from MRI is critical for diagnosis and treatment, but current analysis is time-consuming and challenging due to similar brain tissue characteristics.
- Existing computational methods for brain tumor identification often have limitations, necessitating the development of novel, effective approaches.
Purpose of the Study:
- To present a novel computational method for classifying multiple types of brain tumors non-invasively.
- To enhance the accuracy and efficiency of brain tumor diagnosis using magnetic resonance imaging (MRI) data.
- To address the limitations of traditional biopsy and existing computational techniques in brain tumor classification.
Main Methods:
- A novel brain tumor classification method utilizing MRI data.
- Implementation of the Canny Mayfly segmentation algorithm for image processing.
- Feature selection using the Enhanced chimpanzee optimization algorithm (EChOA) to reduce dimensionality.
- Classification of features using the ResNet-152 model and a softmax classifier.
- The proposed method was implemented in Python using the Figshare dataset.
Main Results:
- The proposed computational method achieved a high accuracy of 98.85% in classifying brain tumors.
- The system demonstrated strong performance in terms of accuracy, specificity, and sensitivity.
- The EChOA effectively minimized the dimensionality of extracted features, aiding classification.
Conclusions:
- The developed non-invasive computational approach offers a promising alternative for accurate and efficient brain tumor classification.
- The integration of Canny Mayfly segmentation, EChOA feature selection, and ResNet-152 classification provides a robust system for brain cancer diagnosis.
- This study highlights the potential of advanced AI techniques in improving outcomes for brain tumor patients by overcoming the challenges of traditional diagnostic methods.
