نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

A radiomics model based on preoperative diffusion-tensor imaging (DTI) for prediction of MGMT promotor methylation status as a clinical biomarker for personalizing treatment strategies in glioma patients

Subject:

Artificial Intelligence and Brain Mapping

Presenter Author:
Name:	Fateme Shahedi

Author:
Name 1:	Fateme Shahedi - Presenter
Affiliation 1:	Department of Radiology Technology, School of Paramedical Sciences, Mashhad University of Medical Sciences, Mashhad, Iran

Name 2:	Narges Mamuri
Affiliation 2:	Medical Physics Research Center, Mashhad University of medical Science, Mashhad, Iran.

Name 3:	Hoda Zare - Corresponding
Affiliation 3:	Medical Physics Research Center, Mashhad University of medical Science, Mashhad, Iran.

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Glioma is the most common primary central nervous system tumor in adults characterized by significant biological heterogeneity, which impacts prognosis and treatment effectiveness. The detection of epigenetic prognostic biomarker such as O6-methylguanine-DNA methyltransferase (MGMT) plays an important role in the prognosis and prediction of treatment response in glioma patients. MGMT is a DNA damage repair enzyme, which decreases the efficacy of the alkylating agent temozolomide that is used in chemotherapy, however, MGMT promoter methylation results in increased response to alkylating agents and longer survival of patients. Therefore, the detection of MGMT methylation status is critical for improving patient prognosis and personalized treatment plans. Currently, the genetic information of adult diffuse gliomas is mainly obtained by sequencing the biopsy, which is invasive, expensive, and time-consuming. The purpose of this study was to investigate the value of radiomics features extracted from the pre-operative conventional structural MRI and diffusion tensor imaging (DTI) in predicting MGMT promoter methylation in patients with diffuse gliomas.

Methods:

We retrospectively analyzed 100 patients from the publicly available UCSF Preoperative Diffuse Glioma MRI dataset. The imaging protocol included conventional MRI sequences (T1WI, T2WI, T1CE, FLAIR) as well as diffusion tensor imaging (DTI)–derived fractional anisotropy (FA) and mean diffusivity (MD) maps. The preprocessing steps included skull stripping, N4 bias field correction to reduce intensity inhomogeneity, noise reduction, and z-score normalization. Radiomic features were extracted using the open-source PyRadiomics package (v3.0). Radiomics features extracted within enhancing regions of tumor tissue, categorize into three classes including shape, first-order statistics, and texture features. Image filtered types applied were wavelet transformation, Laplacian of Gaussian (LoG), and local binary pattern (LBP). The feature selection encompasses three steps: Spearman correlation coefficient, followed by the Mann–Whitney U test to identify discriminative features and implementation of the least absolute shrinkage and selection operator (LASSO) to select the final stable features. Support vector machine (SVM) with a radial basis function (RBF) kernel was used as classifier. Hyperparameters were optimized using Bayesian search within a 5-fold cross-validation repeated five times. To obtain the global optimal performance of the models, a Bayesian optimization algorithm was employed with the aid of 5-fold cross validation to search for the most appropriate hyperparameters of the models.

Results:

The model based on conventional MRI sequences alone achieved an AUC of 0.724. The combined model of DTI-derived fractional anisotropy (FA) and conventional sequences yielded the highest AUC of 0.841. The conventional + MD model achieved the AUC of 0.754. The full multiparametric model (conventional + FA + MD) achieved an AUC of 0.839, which was comparable to the conventional + FA model but superior to both the conventional-only and conventional + MD models.

Conclusion:

Compared with methods relying on a single modality, the combined structural MRI, and DTI -derived FA model demonstrated promising performance in predicting MGMT methylation status, which fully verified the superiority of the multimodal data fusion strategy in improving the prediction accuracy. The established prediction model may help neuro-oncologists to make more precise diagnoses and personalized treatment plans for glioma patients.

Keywords:

Glioma, MGMT promoter methylation, Diffusion-tensor imaging