Dyslexia is a type of learning disability associated with difficulty in reading, despite possessing normal cognitive abilities. Individuals with dyslexia struggle with recognizing and decoding words, spelling, and reading fluency. Our attempt is to use Artificial Intelligence, more specifically Machine learning techniques, to classify patients with dyslexia by analyzing their functional Magnetic Resonance Imaging (fMRI) data. This can lead to an earlier diagnosis of dyslexic patients, which can significantly improve their lives, both academically and socially.
To achieve this, we have employed a total of four models – Support Vector Machine(SVM), Artificial Neural Network(ANN), Convolutional Neural Network(CNN) and a Hybrid CNN-SVM. Moreover,we have employed dimensionality reduction techniques – Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA) – to improve the models’ classification accuracy, before classifying with SVM and ANN. Among these models, the CNN and CNN – SVM model outclassed the other models with an accuracy of 94.4% and 94% respectively. Furthermore, the utilization of PCA resulted in an accuracy of 78% and 89% for SVM and ANN models, respectively, outperforming the use of LDA.
In addition, we explored novel possibilities by employing the Vision Transformer (ViT) Model, known for its high performance in image data. The proposed ViT model was able to surpass our previous models and all existing models with an exceptional accuracy score of 98%. The findings of this study have the potential to inform the development of more accurate and reliable diagnostic tools for dyslexia, leading to more effective therapeutic interventions and ultimately improving the quality of life of individuals affected by the disorder.
Guided by – Adarsh Pradhan
Group Members:
- Abbash Ali
- Achintam Kalita
- Punasmita Ghosh
Branch – CSE
Semester – 8th Sem
'