Predictive value of radiomics modelling based on dynamic and static 18 F-FDG PET/CT imaging for the differential diagnosis of lymph nodes in lung cancer

Xieraili Wumener^1*Zhiheng Hu²Jiuhui Zhao¹Hongjian Wang²Yarong Zhang¹Yaohong Deng³Jun Zhao⁴Ying Liang¹

• ¹National Cancer Center, Cancer Hospital Shenzhen Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Shenzhen, China
• ²The Fourth Affiliated Hospital of Xinjiang Medical University, Xinjiang, China
• ³United Imaging Intelligence Co Ltd, Shanghai, China
• ⁴Tongji University Dongfang Hospital, Shanghai, China

We aimed to identify the most effective machine learning model for predicting the value of the differential diagnosis of lymph nodes (LNs) in lung cancer using dynamic and static 18 F-fluorodeoxyglucose (FDG) positron emission tomography/CT (PET/CT) imaging. Methods: A total of 279 pathologically confirmed LNs from 74 patients with lung cancer were retrospectively analyzed. It was randomly divided into a training group (N = 196) and a test group (N = 83) at a ratio of 7:3. Radiomics features of the images were extracted from CT, dynamic PET (dPET) and static PET (PET) images and were screened for the most predictive value. Build Support Vector Machine (SVM), Logistic Regression (LR) and Random Forest (RF) machine learning models were build using the optimal radiomics features. The best quantitative prediction model was suggested using SUVmax and Ki based on LNs. A composite model was build based on the best machine learning model and quantitative models. Receiver operating characteristic (ROC) curves were used to evaluate the predictive ability of the machine learning, quantitative and composite models to predict LNs metastasis in lung cancer. Results: Of the three machine learning models, the RF model demonstrated the greatest predictive efficacy in both the training (AUC of 0.823) and test (AUC of 0.819) groups. The quantitative model based on Ki showed good predictive efficacy in both the training (AUC of 0.772) and test groups (AUC of 0.805) groups. A composite model based on both the RF machine learning model and the quantitative model demonstrated superior predictive efficacy. The respective AUCs in the training and test groups were 0.844 and 0.835, respectively. Decision curve analysis has shown that the composite model has better net benefit and clinical value. Conclusion: A composite model based on an RF model of PET/CT+Ki images, combined with dynamic quantitative Ki, is highly effective in differentiating FDG-avid LN metastasis in lung cancer. This model provides greater net benefit and clinical value.