ShQDFHNet: Shepard quantum dilated forward harmonic net for brain tumour detection using MRI image

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Healthcare
• Neuro-oncology

Background:

• Brain tumors represent a significant health challenge, with current diagnostic systems primarily relying on medical imaging.
• Accurate and early detection of brain tumors is crucial for effective treatment planning and patient outcomes.

Purpose of the Study:

• To develop and evaluate a novel deep learning model, the Shepard Quantum Dilated Forward Harmonic Net (ShQDFHNet), for enhanced brain tumor detection using MRI scans.
• To improve the accuracy and efficiency of brain tumor segmentation and detection through advanced image processing and feature extraction techniques.

Main Methods:

• Image enhancement using high boost filtering to accentuate critical features.
• Accurate tumor segmentation via Log-Cosh Point-Wise Pyramid Attention Network (Log-Cosh PPANet) with Log-Cosh Dice Loss.
• Extraction of texture features including Spatial Grey-Level Dependence Matrix (SGLDM) and Gray-Level Co-occurrence Matrix (GLCM).
• Brain tumor detection using the ShQDFHNet model, integrating Shepard Convolutional Neural Network (ShCNN) and Quantum Dilated Convolutional Neural Network (QDCNN) with Forward Harmonic Analysis Network layers.

Main Results:

• The ShQDFHNet model demonstrated high performance on the Brain Tumour MRI dataset.
• Achieved an accuracy of 90.69%, a True Positive Rate (TPR) of 91.14%, and a True Negative Rate (TNR) of 90.61% with 9-fold cross-validation.

Conclusions:

• The integration of high boost filtering, Log-Cosh PPANet, and texture-based features significantly enhances input data quality and enables precise tumor segmentation in MRI scans.
• The proposed ShQDFHNet model effectively improves feature learning, leading to robust performance in brain tumor detection from MRI data.