Adult-type diffuse glioma prediction using MnasNet optimized by the advanced single candidate optimizer

Area of Science:

• Neuro-oncology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Diffuse glioma is an aggressive adult primary brain tumor with limited treatment and poor prognosis.
• Current diagnostic methods like biopsy are invasive, time-consuming, and have variable results.
• There is a need for non-invasive, accurate methods for early glioma detection and classification.

Purpose of the Study:

• To develop and validate a non-invasive deep learning approach for predicting adult-type diffuse glioma.
• To utilize preoperative T2-weighted Magnetic Resonance Imaging (MRI) for glioma prediction.
• To enhance diagnostic accuracy and overcome limitations of traditional methods.

Main Methods:

• A deep learning model based on an alternated MnasNet design was employed.
• The model was optimized using the Advanced Single Candidate Optimizer (ASCO) with opposition-based learning and Chebyshev chaotic map.
• The system was trained and validated on a dataset of 533 patients using 10-fold cross-validation.

Main Results:

• The proposed model achieved high performance metrics: 95.11% sensitivity, 96.57% specificity, 98.75% precision, 97.30% accuracy, 97.76% F1-score, and 92.62% MCC.
• Comparative analysis demonstrated superior performance against six state-of-the-art techniques.
• The model accurately predicted glioma status (IDH mutation and 1p/19q codeletion).

Conclusions:

• The developed deep learning system offers an accurate and non-invasive method for predicting adult-type diffuse glioma.
• This approach shows significant clinical potential to improve glioma diagnosis and patient management.
• The AI-driven prediction using T2-weighted MRI represents a significant advancement over conventional diagnostic procedures.