Lung Cancer Classification Using the Shifted Window Transformer (Swin Transformer) Based on Histopathological Images

primadiansyah, dimas aji (2026) Lung Cancer Classification Using the Shifted Window Transformer (Swin Transformer) Based on Histopathological Images. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010306.-cover.pdf Download (750kB)
	Text (Bab 1) 22081010306.-bab1.pdf Download (129kB)
	Text (Bab 2) 22081010306.-bab2.pdf Restricted to Repository staff only until 22 July 2029. Download (704kB)
	Text (Bab 3) 22081010306.-bab3.pdf Restricted to Repository staff only until 22 July 2029. Download (1MB)
	Text (Bab 4) 22081010306.-bab4.pdf Restricted to Repository staff only until 22 July 2029. Download (874kB)
	Text (Bab 5) 22081010306.-bab5.pdf Download (115kB)
	Text (Daftar Pustaka) 22081010306.-daftarpustaka.pdf Download (116kB)

Abstract

highlighting the need for a system that can assist the diagnostic process quickly and accurately. This study aims to implement the Shifted Windows Transformer (Swin Transformer) method for classifying lung cancer histopathological images using the LC25000 dataset, which consists of 15,000 images divided into three classes: Lung Adenocarcinoma, Lung Benign Tissue, and Lung Squamous Cell Carcinoma. The research stages include image preprocessing, hyperparameter tuning using Grid Search, model training, evaluation using accuracy, precision, recall, F1-score, confusion matrix, and ROC curve, as well as model deployment into a web-based application using Laravel as the frontend and Flask as the backend. The experimental results show that the Swin Transformer model achieved an accuracy of 98.13% on the testing dataset and 85.00% on blind testing using the LungHist700 dataset, outperforming both the Vision Transformer (ViT) and ResNet50 models. These findings demonstrate that the Swin Transformer is effective for lung cancer classification based on histopathological images and has the potential to serve as a computer-aided diagnostic support system.

Item Type:

Thesis (Undergraduate)

Contributors: