DyslexiaAnalyzer: Effective Dyslexia Handwriting Analysis using Transfer Learning with Explainable AI

Dyslexia is a very common learning disability impacting reading, writing, and spelling, remains unreported for its subtle signs of handwriting. This motivates for the early diagnosis in order to help the children intervene and receive assistance. This paper focuses on "DyslexiaAnalyzer" the latest innovation for the detection of dyslexia in handwriting that employing transfer learning with Explainable AI (XAI). This work uses MobileNetV3-a lightweight but robust pretrained convolutional neural network to classify handwritings. To ensure transparency and interpretability, Grad-CAM (Gradient-weighted Class Activation Mapping) is integrated, offering visual insights into the model’s decision-making process. The proposed model achieves a remarkable classification accuracy of 96.54% that demonstrates its effectiveness and reliability. This high performance underscores the potential of transfer learning in handling complex handwriting datasets with limited samples. Further, the incorporation of XAI techniques addresses the significant challenges in model interpretability that would allow educators and clinicians to trust and embrace AI-driven solutions in sensitive diagnostic contexts. The findings have implications for developing accessible, scalable, and interpretable tools for screening dyslexia. Future research directions should be directed toward increasing dataset diversity, enhancing real-time detection capabilities, and exploring multimodal approaches that involve audio and visual cues. By developing AI-driven diagnostic frameworks, DyslexiaAnalyzer establishes a basis for more inclusive and supportive educational environments.