Annotation Facial Images for Stroke Classification Acute vs Non Acute

This paper presents the annotation and labelling techniques of a dataset employed in deep learning project for the detection and classification of acute stroke and non-stroke in medical images. The dataset comprises 3745 images categorized into two classes: individuals diagnosed with acute stroke and those without stroke. To enhance the dataset's diversity and robustness, we applied various data augmentation methods, including image flipping, rotation, and scaling. These techniques facilitated the creation of a versatile dataset that more accurately mirrors real-world scenarios. This dataset serves as a valuable resource for researchers and healthcare professionals in the stroke medicine domain, offering a comprehensive and varied collection of images for training machine learning models in stroke detection and diagnosis. YOLOv9 models have been utilized to evaluate the object detection capabilities for stroke and non-stroke conditions within the proposed dataset. The dataset and associated annotations are made publicly available for academic use.