Fig. 1
From: MCI detection from handwritten drawing test using residual vision transformer
Overview of the proposed ResViT-based framework for MCI detection. The process begins with (1) input grayscale images from three cognitive drawing tasks: Clock Drawing Test, Cube Copying, and Trail Making Test. (2) Preprocessing involves resizing, normalization, and transformation. (3) The dataset is then split into training, validation, and testing sets. (4) Data augmentation is applied to increase model generalizability. (5a, 5b) Feature extraction is performed in parallel using pre-trained ResNet50 for local features and Vision Transformer (ViT) for global context. (6) Extracted features are fused. (7) The fused features are passed through fully connected layers for classification. (8) The model outputs a prediction indicating either a Healthy or MCI (Mild Cognitive Impairment) status.
