Related Experiment Video
Updated: Jun 4, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Tensor enhanced chest cancer classification via CNN and Vision Transformer models
Nayab Asim1, Mehreen Sirshar1, Mohammad Zubair Khan2
1Software Engineering, Fatima Jinnah Women University, Rawalpindi, Punjab, Pakistan.
Plos One
|June 2, 2026
Summary
This study compares deep learning models for lung cancer detection using CT/PET-CT scans. ResNet-50 and EfficientNet showed high accuracy, while Vision Transformer captured complex patterns, highlighting AI
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Oncology
- Deep Learning for Disease Detection
Background:
- Lung cancer is a major global health concern, necessitating improved diagnostic tools.
- Early and accurate diagnosis of lung cancer is crucial for effective patient treatment and survival rates.
- Current diagnostic methods can be enhanced by advanced computational approaches.
Purpose of the Study:
- To evaluate and compare various convolutional neural network (CNN) architectures against a Vision Transformer (ViT) model for lung cancer classification.
- To introduce a unified platform for preprocessing medical images using a tensor-based pipeline for deep learning models.
- To assess the performance of different deep learning models on the YOLOTransfer dataset for CT/PET-CT image classification.
Main Methods:
- A common tensor-based preprocessing pipeline was applied to all input CT/PET-CT images.
- Convolutional neural network models (AlexNet, VGG-16, ResNet-50, DenseNet, EfficientNet) were compared with a Vision Transformer (ViT) model.
- Model performance was evaluated using metrics including accuracy, sensitivity, specificity, F1-score, and AUC-ROC on the YOLOTransfer dataset.
Main Results:
- ResNet-50 and EfficientNet demonstrated the highest accuracy in lung cancer classification.
- The Vision Transformer model exhibited competitive performance, particularly in identifying complex global patterns within the images.
- The tensor-based preprocessing pipeline facilitated implicit fine-tuning across different model architectures.
Conclusions:
- Deep learning models, including CNNs and ViTs, show significant potential for accurate lung cancer detection from CT/PET-CT imaging.
- The study highlights the complementary strengths of convolutional and transformer-based architectures in medical image analysis.
- The findings support the feasibility and effectiveness of AI-driven approaches for improving early lung cancer diagnosis.