Preoperative Discrimination of CDKN2A/B Homozygous Deletion Status in Isocitrate Dehydrogenase-Mutant Astrocytoma: A...

Area of Science:

• Neuroradiology
• Oncology
• Medical Imaging Analysis

Background:

• Cyclin-dependent kinase inhibitor 2A/B (CDKN2A/B) homozygous deletion is a critical negative prognostic biomarker in IDH-mutant astrocytoma.
• Accurate, noninvasive discrimination of CDKN2A/B deletion status is vital for clinical management.

Purpose of the Study:

• To develop a noninvasive, preoperative model using MR imaging features.
• To discriminate CDKN2A/B homozygous deletion status in IDH-mutant astrocytoma.

Main Methods:

• Retrospective analysis of 251 patients with IDH-mutant astrocytoma.
• Extraction of radiomics and deep learning features from CE-T1WI and T2FLAIR sequences.
• Development and comparison of radiomics, deep learning-based radiomics (DLR), and integrated models.

Main Results:

• The integrated model combining radiomics and DLR features achieved superior performance.
• The final combined model demonstrated high AUC values (test group: 0.943).
• Both radiomics and DLR models showed promising results individually.

Conclusions:

• An integrated deep learning and radiomics model effectively predicts CDKN2A/B homozygous deletion status.
• This noninvasive approach enhances preoperative discrimination for IDH-mutant astrocytoma.
• The combined model offers improved accuracy over models using single feature types.