Thyroid Cancer Detection using Vision Transformers

Engineering thesis project focused on detecting thyroid cancer from ultrasound (USG) images using deep learning techniques.

Technologies Used

• Python
• PyTorch
• TensorFlow
• Scikit-learn
• Pandas, NumPy

Features

• Implemented Vision Transformer (ViT) model for classification.
• Compared performance with traditional Convolutional Neural Networks (CNN).
• Preprocessing and augmentation of medical imaging data.
• Model evaluation based on metrics such as accuracy, precision, recall, and F1-score.

Training data sources

Algerian USG Image Dataset:

https://www.kaggle.com/datasets/azouzmaroua/algeria-ultrasound-images-thyroid-dataset-auitd

USG Dataset for Deep Learning:

https://figshare.com/articles/dataset/An_ultrasonography_of_thyroid_nodules_dataset_with_pathological_diagnosis_annotation_for_deep_learning/26067475?file=47150848