Imaging phenotyping using ¹⁸F-FDG PET/CT radiomics to predict micropapillary and solid pattern in lung adenocarcinoma

Area of Science:

• Oncology
• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung adenocarcinoma is a major cause of cancer-related mortality.
• Micropapillary and solid (MP/S) components are associated with poor prognosis in lung adenocarcinoma.
• Accurate preoperative identification of MP/S components is crucial for treatment planning.

Purpose of the Study:

• To develop and validate a machine learning model for predicting MP/S components in lung adenocarcinoma.
• To integrate 18F-FDG PET/CT radiomics with clinical features for enhanced predictive accuracy.

Main Methods:

• Utilized data from 846 patients with pathologically confirmed adenocarcinoma.
• Extracted and selected 1688 radiomics features from preoperative 18F-FDG PET/CT scans.
• Developed and validated predictive models, including a nomogram combining radiomics and clinical data, using ROC, calibration curves, and DCA.

Main Results:

• The combined PET/CT radiomics-clinical model achieved an AUC of 0.87, outperforming PET (0.829) and CT (0.827) radiomics models in the training cohort.
• The combined model demonstrated superior performance in independent test cohorts.
• Calibration curves showed good agreement, and DCA indicated the highest clinical benefit for the combined model.

Conclusions:

• 18F-FDG PET/CT radiomics signatures show promise in predicting MP/S components in adenocarcinoma.
• The integrated radiomics-clinical model offers the best performance for identifying MP/S components.
• This approach can assist clinicians in tailoring treatment and surveillance strategies for lung adenocarcinoma patients.