Improving prediction accuracy of spread through air spaces in clinical-stage T1N0 lung adenocarcinoma using computed tomography imaging models

Shihua Dou ^1,2,3, Zhuofeng Li ⁴, Zhenbin Qiu ^5,6,7

⁰Second Clinical Medical College, Jinan University, Shenzhen, China.

Jtcvs Open +

|

November 13, 2024

View abstract on PubMed

Summary

Computed tomography (CT) models accurately predict spread through air spaces (STAS) in early lung cancer. These radiomic and deep learning models can guide treatment decisions for T1N0 lung adenocarcinoma patients.

Area Of Science

Medical Imaging
Oncology
Artificial Intelligence

Background

Spread Through Air Spaces (STAS) is a key factor in lung adenocarcinoma recurrence.
Accurate prediction of STAS is crucial for effective surgical planning in early-stage lung cancer.

Purpose Of The Study

To develop and validate computed tomography (CT)-based radiomic and deep neural network (DNN) models.
To enhance the prediction accuracy of STAS in clinical-stage T1N0 lung adenocarcinoma.

Main Methods

Retrospective analysis of 1258 patients with stage T1N0 lung adenocarcinoma.
Development of radiomic models using PyRadiomics and least absolute shrinkage and selection operator regression.
Construction of a two-stage deep learning model (MultiCL) utilizing supervised contrastive learning and fine-tuning.

Main Results

The best radiomic model achieved an area under the curve (AUC) of 0.8944 on the test cohort and 0.7796 on the external validation cohort.
The MultiCL deep learning model achieved an AUC of 0.8434 on the test cohort and 0.7686 on the external validation cohort.
Both model types demonstrated significant predictive ability for STAS status.

Conclusions

CT-based radiomic and DNN models can accurately identify STAS status in T1N0 lung adenocarcinoma.
These imaging models have the potential to improve surgical decision-making.
Improved prediction of STAS can lead to better patient outcomes in lung cancer treatment.

Keywords:

clinical-stage T1N0 lung adenocarcinoma deep neural network radiomics spread through air spaces