.png){kind=icon}
.png){kind=icon}
Return
Developing an interpretable decision support framework for brain tumor screening
.png){kind=icon}
PRE
AI
DOI：10.1007/s10479-026-07262-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Brain and other Central Nervous System (CNS) tumors affect nearly one million people each year, with an estimated 93,470 diagnosed new cases in the United States alone in 2022. There is a significant global imbalance between the supply of medical equipment and the demand for experienced healthcare professionals, especially in developing countries. As a result, experts advocate for innovative technologies to enhance patient screening. This paper presents a comprehensive framework based on eXplainable Artificial Intelligence (XAI) aimed at improving brain tumor screening. Unlike traditional AI models, which often operate as "black boxes" with little transparency, our framework offers accountability through an explainable and interpretable solution, thereby increasing trust among clinical radiologists and oncologists. Key features of our findings include: (1) utilizing an XAI-based method that not only predicts brain tumors but also explains the features that contributed to each prediction; (2) employing three advanced XAI techniques—SHAP, Guided BackPropagation (GBP), and Occlusion Sensitivity (OS)—to create a comparative, multi-class explanatory visual strategy for interpreting CNN-based models; and (3) generating full-color diagnostic images that highlight the volume and severity of each tumor type. Using magnetic resonance imaging (MRI) data from patients, we demonstrate that our XAI-based approach significantly outperforms existing state-of-the-art AI methods for brain tumor screening, ultimately improving clinical decision-making.
Keywords:
Explainable artificial intelligence
Computational design science
Brain tumor
Deep learning
Health screening
Journal
IF:
4.5
Papers:
8.0K
Citations:
2.1W