Brain tumor detection using hybrid transfer learning and patch antenna-enhanced microwave imaging

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Medical Imaging

Background:

• Brain tumors present a significant challenge in healthcare, requiring early detection and precise monitoring for effective treatment.
• Current diagnostic methods like MRI and CT scans have limitations.
• There is a need for innovative, non-invasive diagnostic techniques.

Purpose of the Study:

• To propose an innovative technique integrating hybrid transfer learning with improved microwave imaging for brain tumor detection.
• To leverage pre-trained deep learning models for feature extraction and patch antennas for high-resolution imaging.
• To develop an AI-based detection model for effective brain tumor classification.

Main Methods:

• Development of a patch antenna and head phantom model for SAR analysis and feature extraction.
• Implementation of an AI detection model using MobileNet V2 for image analysis.
• Utilizing depth-wise separable convolutions and inverted residual blocks in MobileNet V2 for high-level feature extraction.
• Classification of brain tumors using a fully connected layer with extracted features.

Main Results:

• The developed model demonstrated exceptional performance in simulations.
• Achieved an accuracy of 98.44%, precision of 98.03%, recall of 99.00%, F1-score of 98.52%, and specificity of 97.82%.
• The hybrid transfer learning approach proved effective for brain tumor detection.

Conclusions:

• The proposed method offers a promising solution for non-invasive, real-time brain tumor detection.
• This technique utilizes the electromagnetic properties of brain tissue and AI capabilities.
• It addresses limitations associated with conventional diagnostic methods like MRI and CT scans.