Abstract:
Lung diseases are a leading cause of death worldwide, and early detection is crucial for
effective treatment. Two commonly used imaging techniques for detecting lung diseases are
Computed Tomography (CT) and X-rays. However, interpreting these scans can be timeconsuming and subjective, requiring significant expertise and experience. Deep Learning
techniques have shown potential in automating the detection and diagnosis of lung diseases
through CT and X-ray scans. This approach involves training models on large datasets of scans
and associated disease labels, allowing the models to learn and identify patterns and features
indicative of various lung diseases. By using these models, clinicians can obtain more accurate
and efficient diagnoses, which leads to improved patient outcomes. These techniques can
potentially revolutionize lung disease detection and diagnosis, making them more efficient,
accurate, and accessible. In this work, the automatic detection of lung disease named COVID-19
is performed using Deep Learning-based Vision Transformers. This research is conducted on the
CT Scans dataset. The pre-trained model ViT-Base-Patch16-224-in21k on the ImageNet dataset
is used. Data augmentation and transfer learning are applied to our CT-Scan dataset to attain
higher accuracy. The proposed model has achieved the highest training and validation accuracies,
which are 99.65% and 99.20% respectively.
