Prediction of IDH and 1p/19q Status in Gliomas Based on Dual Structural Feature Exploration and Alignment Network

Area of Science:

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

Background:

• Accurate preoperative prediction of isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion status in gliomas is crucial for patient prognosis and treatment strategies.
• Glioma heterogeneity presents a significant challenge for machine learning models attempting to predict these statuses from multisequence magnetic resonance imaging (MRI) data.

Purpose of the Study:

• To develop and validate a novel deep learning framework, the dual structural feature exploration and alignment network (DSFEAnet), for noninvasive prediction of IDH mutation and 1p/19q codeletion status in gliomas using preoperative MRI.
• To effectively capture and leverage both intratumoral and marginal glioma heterogeneity from MRI images to improve prediction accuracy.

Main Methods:

• Proposed DSFEAnet incorporates a match and mismatch feature extraction (MMFE) module to identify intratumoral heterogeneity features (e.g., tumor core localization, T2-FLAIR mismatch).
• Integrated a graph-based geometry exploration (GGE) module to analyze marginal heterogeneity by representing tumor surface variations as a graph.
• Employed a dual structural feature alignment (DSFA) module to fuse and align extracted intra- and inter-structural features, enhancing overall representational power.

Main Results:

• DSFEAnet demonstrated robust performance on multicenter and independent clinical datasets.
• Achieved an Area Under the Curve (AUC) of 87.72% for IDH mutation status prediction and 80.52% for 1p/19q codeletion status prediction on the Nanfang hospital dataset.
• Interpretability analysis confirmed the effectiveness of the proposed MMFE and GGE modules in capturing relevant glioma features.

Conclusions:

• The DSFEAnet model shows significant potential for accurate, noninvasive preoperative prediction of IDH mutation and 1p/19q codeletion status in gliomas.
• This approach can aid clinicians in making more informed decisions regarding glioma patient management and therapeutic planning.