Related Experiment Video
Updated: May 8, 2026
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.1K
A position-enhanced sequential feature encoding model for lung infections and lymphoma classification on CT images
Rui Zhao1,2, Wenhao Li3, Xilai Chen4
1Research Center for Medical AI, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
Summary
This study introduces a new AI model for lung CT scans to better distinguish pulmonary lymphoma from infections. The transformer-based approach improves accuracy in classifying these challenging lung conditions.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer-Aided Diagnosis
Background:
- Distinguishing pulmonary lymphoma from lung infections on CT scans is difficult.
- Current deep learning models using 2D slices miss comprehensive information, while 3D models face performance limitations due to information compromise and parameter reduction.
Purpose of the Study:
- To develop an advanced deep learning model for accurate classification of lung infections and pulmonary lymphoma using complete 3D CT images.
- To overcome the limitations of existing 2D and 3D CT classification models.
Main Methods:
- Proposed a transformer sequential feature encoding structure integrating multi-level information from complete CT images.
- Incorporated position encoding and cross-level long-range information fusion modules within a CNN network for enhanced feature extraction.
Main Results:
- The model achieved high performance on a dataset of 124 patients.
- Demonstrated superior accuracy (0.875), AUC (0.953), and F1 score (0.889) compared to state-of-the-art methods in distinguishing lung infections from pulmonary lymphoma.
Conclusions:
- The position-enhanced transformer-based sequential feature encoding model effectively extracts high-precision and contextual features in lung CT images.
- The proposed method significantly improves classification performance for differentiating lung infections and pulmonary lymphoma.
- Source code is publicly available.
