Integrated predictive model for visceral pleural invasion in small NSCLC with high clinical utility.
Authors
Affiliations (9)
Affiliations (9)
- Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China.
- Shanghai Institute of Medical Imaging, Shanghai, China.
- Department of Research and Development, United Imaging Intelligence, Shanghai, China.
- Department of Radiology, Shanghai Public Health Clinical Center, Fudan University, Shanghai, China.
- Department of Research and Development, United Imaging Intelligence, Shanghai, China. [email protected].
- Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China. [email protected].
- Shanghai Institute of Medical Imaging, Shanghai, China. [email protected].
- Shanghai Institute of Medical Imaging, Shanghai, China. [email protected].
- Department of Radiology, Shanghai Public Health Clinical Center, Fudan University, Shanghai, China. [email protected].
Abstract
This study aims to develop and validate a multi-feature integrated imaging fusion (MIIF) model, incorporating deep learning, radiomics features, and computed tomography (CT) findings, for identifying visceral pleural invasion (VPI) in small non-small cell lung cancer (NSCLC). This multi-center retrospective analysis included 2822 small NSCLCs. These were divided into four datasets (training, validation, internal/external test). The MIIF model's diagnostic performance was compared against the assessments of six radiologists. Additionally, we evaluated the clinical utility of the MIIF model by comparing the diagnostic performance of radiologists, with/without the aid of the model. The MIIF model yielded AUCs of 0.869/0.785 in the internal/external test sets, respectively, which were comparable to the radiologists' (P > 0.05). With MIIF assistance, radiologists' accuracy and specificity increased to 0.845/0.828 and 0.836/0.841 in the internal/external test sets (P < 0.001). The MIIF model shows enhanced accuracy and specificity in detecting VPI in small NSCLC and may improve radiologist' diagnostic performance.