An eXplainable deep learning model for multi-modal MRI grading of IDH-mutant astrocytomas

IDH-Mutant-astrocytomas are malignant brain glioma tumors. They are graded as lower grade (grades 2 and 3) or higher grade (grade 4) according to their rate of growth and molecular features. Conventional IDH grading requires pathology examination, which is invasive, costly, and time-consuming. Advanced MRI modalities are widely used to overcome some of these limitations but require a high level of interpretation for tumor grading. Recent advances in deep learning have shown potential to aid and improve the accuracy and efficiency of tumor classification. This study proposes a novel non-invasive approach for grading IDH-Mutant-astrocytomas using a Light-weight Attention Network (LAN) that integrates multi-modal MRI data, including both conventional and advanced MRI modalities. The study addresses inter-modality heterogeneity using Principal Component Analysis (PCA) while minimizing computational complexity. LAN uses a 3D Convolutional Neural Network (CNN) and a volumetric attention mechanism to extract tumor patterns and classify grades 2 to 4. This is followed by an eXplainable AI approach that uses SHapley Additive exPlanations (SHAP) to interpret model decisions and identify key contributing features. The proposed LAN model outperforms other pre-trained state-of-the-art models, achieving an overall accuracy of 0.84. The SHAP attribution scores demonstrate that advanced MRI modalities such as Arterial Spin Labeling (ASL) and Diffusion-weighted Imaging (DWI), along with conventional MRI sequences such as FLAIR, T1-c and T2, contribute significantly to improving tumor heterogeneity visualization for aggressive grade 2 gliomas. This work demonstrates the feasibility of integrating explainable deep learning with multi-modal MRI data for precise and comprehensible early glioma grading, potentially leading to better clinical decision-making. • Introduces a multimodal Light-Attention Network for grading IDH mutant astrocytomas. • Integrates traditional and advanced MRI modalities for enhanced analysis. • Applies PCA to minimise inter-modality heterogeneity. • Utilises 3D Convolution and attention mechanism for feature extraction. • SHAP explains MRI modality attributions, with ASL and DWI significant in grading.