XMorph: Explainable Brain Tumor Analysis Via LLM-Assisted Hybrid Deep Intelligence

Explainability and accuracy are compatible in medical AI: XMorph achieves 96.0% classification accuracy for brain tumors while remaining interpretable, challenging the assumption that high-performing AI systems must be "black boxes" that clinicians cannot understand.

Hybrid explanation approach improves clinical trust: The framework combines visual explanations (GradCAM++) with AI-generated text descriptions, providing doctors with multiple ways to understand why the system made a particular diagnosis—addressing a key barrier to clinical adoption.

Focusing on tumor boundaries matters: The Information-Weighted Boundary Normalization mechanism specifically emphasizes irregular tumor edges, which are clinically important for distinguishing malignant from benign tumors, demonstrating that AI design choices should reflect medical expertise.

Computational efficiency enables broader deployment: By reducing computational demands alongside improving interpretability, XMorph addresses practical barriers to implementing AI in clinical settings, particularly in resource-limited healthcare environments.

Open science supports responsible AI development: The authors' commitment to publicly sharing code and materials enables independent verification, reproducibility, and broader scrutiny—essential practices for responsible deployment of AI in healthcare.